epigenetic study of colorectal cancer: lncrnas and cimp ... · biologia do cancro, do ipo-porto. a...

Epigenetic Study of Colorectal Cancer: lncRNAs and CIMP Profiling.Fábio Miguel Tavares FerreiraMaster Degree in BiochemistryChemistry and Biochemistry Department2016

SupervisorsCarmen Jerónimo, PhD,Guest Associate Professor with Habilitation at ICBASAssistant Investigator and Coordinator of the Cancer Biology and Epigenetics Group at IPO-Porto

Pavel Vodička, MD, PhD, Senior Scientist and Coordinator of the Department of the Molecular Biology of Cancer at IEM ASCR, v.v.i.

Co-SupervisorAlena Opattová, PhD, Post-Doctoral Researcher at IEM ASCR, v.v.i.

Todas as correções determinadas pelo júri, e só essas, foram efetuadas.

O Presidente do Júri,

Porto, ______/______/_________

“And so gentlemen, I learned. Oh, if you have to learn, you

learn; if you’re desperate for a way out, you learn; you learn

pitilessly. You stand over yourself with a whip in your hand;

if there’s the least resistance, you lash yourself.”

― Franz Kafka, The Metamorphosis and Other Stories

FCUP Epigenetic Study of Colorectal Cancer: lncRNAs and CIMP Profiling.

I

AGRADECIMENTOS Agradeço a todos aqueles que, direta ou indiretamente, contribuíram não só para a

elaboração desta Dissertação, mas também para a experiência que tive e conhecimento que

adquiri ao realizar estes dois projetos. Um enorme obrigado dedicado em particular às

pessoas que me motivaram e auxiliaram durante a experiência de ERASMUS+Estágio.

Começo por agradecer aos membros da Comissão Científica do Mestrado pela

disponibilidade e o apoio a que se predispuseram, não só a mim, mas a todos os meus

colegas. Principalmente ao meu tutor, Professor Pedro Alexandrino, pelas muitas palavras

de apoio e pelos valiosos conselhos que foram chave nos momentos mais difíceis.

I’d like to thank my supervisor in Prague, Dr. Pavel Vodička, and Dr. Ludmila

Vodičková for their kindness and comprehension, and for having accepted me in their lab,

otherwise I would never live the bright side of the experience that Erasmus represented. A

special “thank you” to Dr. Ludmila Vodičková for being always worried about my situation

and ready to help me out. In addition, I’d also like to thank all the people in the laboratory for

being constantly so friendly, honest, worried and funny. To Alena and Andrea, I owe all the

accomplished in the first project exposed herein, and all that I have learned from that. My

special gratitude is for them.

À minha orientadora, Professora Cármen Jerónimo, pelo enorme profissionalismo,

disponibilidade, orientação e ensinamentos que me transmitiu neste estágio que

proporcionou e no qual me integrou. É sobretudo a ela que se deve a existência desta

Dissertação e a realização do segundo e maior projeto aqui apresentado. Pela paciência e

compressão, pelas críticas, mas também pela confiança. Por tudo, um franco Obrigado.

Gostaria também de agradecer aos restantes membros do grupo de Epigenética &

Biologia do Cancro, do IPO-Porto. A todos eles estou grato por proporcionarem um ambiente

profissional, mas também bastante agradável, pela simpatia e pelo espírito de entreajuda e

integração. Um especial agradecimento à Micaela Freitas e à Catarina Barbosa pela sua

enorme contribuição, sacrifício, simpatia e instrução; e ao Eng. Luís Antunes do Serviço de

Epidemiologia IPO-Porto por toda a valiosa ajuda estatística.

Finalmente, agradeço à minha família pelo apoio sempre presente: aos meus

pais (Maria Olinda e Fernando) e aos meus avós (Arlindo, Cipriano, Júlia e Leopoldina), e

especialmente às minhas irmãs Cláudia e Marlene (que fez um grande esforço). A eles

dedico esta Dissertação. Sem dúvida que também os meus amigos foram o motor de todo

este processo. Um especial obrigado ao Ricardo, à Joana Marques, à Joana, à Rita e à

Susana. Também à Verónica e à Ana Freitas, à Bárbara, ao Nuno, ao Henrique, ao Tiago e

ao Márcio, e sobretudo ao Carlos. A todos os meus restantes amigos e colegas de curso

também dedico o meu apreço.

Este estudo foi parcialmente financiado por uma bolsa do Centro de Investigação do

Instituto Português de Oncologia do Porto (CI-IPOP-74-2016).

II FCUP Epigenetic Study of Colorectal Cancer: lncRNAs and CIMP Profiling.


III

ABSTRACT PROJECT I: Role of lncRNAs in the regulation of DNA repair.

DNA damage is a lethal and common event during the lifetime of a cell, and in

view of its repair, some specific pathways have evolved, integrating the general DNA

damage response (DDR). As expected, DNA repair alterations have been extensively

correlated with cancer; and in particular colorectal cancer (CRC), which often presents

genome instability due to DNA mismatch repair (MMR) deficiency. In contrast with MMR,

the role played by other DNA repair pathways in CRC are not so well reviewed. In a

comprehensive study by Slyskova et al (2012)1, base and nucleotide excision repair

pathways (BER and NER, respectively) were found not to be considerably altered in

CRC. However, to further investigate the possible involvement of excision repair in CRC,

an epigenetic analysis considering the biggest and less studied class of transcripts was

proposed. Long non-coding RNAs (lncRNAs) are a miscellaneous class of multi-

functional RNA molecules that has been recently correlated with CRC and also with

DDR. Therefore, the purpose of this project was the discovery of BER-related lncRNAs,

which could represent new biomarkers or treatment-targets in CRC.

The same CRC tissue samples from Slyskova et al (2012) were used along with

the obtained data to generate four distinct groups with five elements each, divided

according to lower and higher DNA repair capacity (DRC) measurements in both cancer

and adjacent healthy tissues. Using a LncProfiler qPCR Array®, the levels of ninety

lncRNAs were measured for each of the twenty selected samples, and next analysed for

statistically significant expression differences.

This analysis revealed the inexistence of significant differences: neither between

each pair of groups compared, nor between all tumour versus normal mucosa samples

or lower DRC versus higher DRC; due to the small size of the series and high inter-

variability. Hence, although these results indicate that no possible role exists for the

tested lncRNAs in CRC tumourigenesis in association with BER functionality, no solid

conclusions can be stated.

PROJECT II: Evaluation of CIMP status in colorectal cancer and correlation with prognosis.

Colorectal cancer (CRC) is one of the major causes of cancer-related morbidity

and mortality worldwide. Despite of recent advances in treatment approaches, cancer

IV FCUP Epigenetic Study of Colorectal Cancer: lncRNAs and CIMP Profiling.

progression and metastization still remains a major concern. This heterogeneous

disease is currently classified according to global genomic or epigenomic status, which

have been linked to different clinicopathologic characteristics, prognosis and treatment

response. Therefore, segregation of CRC patients by their molecular phenotype is

essential to predict those who will benefit from a specific therapy. A subset of CRC

patients has been shown to exhibit widespread promoter CpG island methylation, termed

CpG Island Methylator Phenotype (CIMP). For instance, CIMP has been increasingly

referred as a promising prognostic factor. However, it is the less understood molecular

subtype in CRC and various methods and definitions have been used to categorize CIMP

status, leading to discrepancies. To further assess this issue, new integrative studies are

required. Thus, the main goal of this project was CIMP profiling and the analysis of a

series of 211 patients diagnosed with sporadic CRC.

DNA extracted from 211 CRC and 43 healthy mucosa samples, formalin fixed

paraffin embedded, was bisulfite converted, and promoter methylation of five genes/loci

was then assessed by real-time qMSP (SYBR® Green-based), for CIMP frequency.

Further statistical analysis to disclose associations with clinicopathological parameters,

and survival analyses to evaluate CIMP prognostic value were conducted.

CIMP was found in 8.5% of all CRC cases and did not associate with any of the

studied clinicopathological and molecular variables. Furthermore, CIMP did not

associate with patients’ prognosis, both for disease-specific survival (DSS) (HR 1.192

95% CI 0.732-1.941, P=0.481) or disease-free-survival (DFS) (HR 0.554 95% CI 0.241-

1.275, P=0.161). However, aberrant methylation of one of the five markers constituting

the selected panel, CDKN2A(p16), associated with shorter DSS, but only in univariable

analysis (HR 1.578 95% CI 1.016-2.450, P=0.042).

CIMP status did not associate with patients’ survival, which is in accordance with

previous studies by others. However, the laboratory technique or its application with the

specific panel selected may not be adequate to evaluate CIMP status, yielding lower

CIMP frequencies and further lack of significant associations between CIMP and any of

the recorded variables. Additional studies are needed to further confirm these preliminary

results.

Keywords: Epigenetics, colorectal cancer, lncRNAs, base-excision repair, methylation,

CIMP, prognosis.


V

RESUMO

PROJETO I: Papel dos lncRNAs na regulação da reparação do ADN.

Os danos no ADN são um evento letal e comum durante o tempo de vida de uma

célula, e tendo em vista a sua reparação, algumas vias específicas evoluíram,

integrando no geral a resposta a danos no ADN (DNA damage response – DDR). Como

esperado, alterações na reparação do ADN têm sido extensivamente correlacionadas

com o cancro; e, em particular o cancro colorectal (CCR), que muitas vezes apresenta

instabilidade genómica devido a defeitos na via de reparação de desemparelhamentos

(mismatch repair – MMR). Em contraste com a via MMR, o papel desempenhado por

outros mecanismos de reparação do DNA no CCR não está tão bem revisto. Num

estudo abrangente por Slyskova et al (2012)1, as vias de reparação por excisão de bases

ou nucleótidos (base-excision repair – BER e nucleotide-excision repair – NER,

respetivamente) não foram consideradas notavelmente alteradas no CCR. No entanto,

para investigar o possível envolvimento da reparação por excisão no CCR, foi proposta

uma análise epigenética tendo em conta a maior e menos estudada classe de

transcritos. ARNs não-codificantes longos (long non-coding RNAs – lncRNAs) são uma

classe variada de moléculas de ARN multifuncionais que foi recentemente

correlacionada com o CCR e também com DDR. Portanto, o objetivo deste projeto foi a

descoberta de lncRNAs relacionadas com a via BER, que poderão representar novos

biomarcadores ou alvos de tratamento para o CCR.

As mesmas amostras de tecido de CCR estudadas em Slyskova et al (2012),

juntamente com a respetiva informação obtida, foram utilizadas para criar quatro grupos

distintos, com cinco elementos cada, divididos de acordo com uma menor ou maior

capacidade de reparação do ADN (DNA repair capacity – DRC) – determinada tanto no

tecido tumoral como da mucosa normal adjacente. Usando LncProfiler qPCR Array®,

os níveis de noventa lncRNAs foram medidos para cada uma das vinte amostras

selecionadas, e em seguida analisados relativamente à existência de diferenças

estatisticamente significativas na expressão.

Esta análise revelou a inexistência de diferenças significativas: nem entre cada

par de grupos comparados, nem entre todas as amostras tumorais versus amostras de

mucosa normal, ou menor DRC versus maior DRC, devido ao reduzido tamanho

amostral e à elevada inter-variabilidade. Assim, embora estes resultados indiquem que

não existe qualquer papel para os lncRNAs testados na tumorigénese do CCR em

VI FCUP Epigenetic Study of Colorectal Cancer: lncRNAs and CIMP Profiling.

associação com a funcionalidade da via BER, não podem ser apontadas conclusões

sólidas.

PROJECT II: Avaliação do perfil CIMP no cancro colorectal e correlação com prognóstico.

O cancro colorectal (CCR) é uma das principais causas de morbidade e

mortalidade relativas a cancro no mundo. Apesar dos recentes avanços de abordagens

terapêuticas, progressão do cancro e metastização ainda persistem como a principal

preocupação. Esta doença heterogênea é atualmente classificada em função do estado

genético e epigenético global, o que tem sido associado com diferentes características

clinicopatológicas, prognóstico e tratamento. Assim, a segregação de pacientes com

CCR pelo seu fenótipo molecular é essencial para prever aqueles que irão beneficiar de

uma terapia específica. Um subconjunto de pacientes com CCR demonstrou exibir

metilação generalizada em ilhas CpG de promotores, o que foi denominado Fenótipo

Metilador de Ilhas CpG (CpG Island Methylator Phenotype – CIMP). Efetivamente, CIMP

tem sido, cada vez mais, referido como um promissor fator de prognóstico. No entanto,

é o subtipo molecular menos compreendido no CCR, e várias definições e métodos têm

sido utilizados para categorizar o perfil CIMP, conduzindo a discrepâncias. Para avaliar

mais profundadamente esta questão, novos estudos integrativos são necessários.

Assim, o objetivo principal deste projeto foi o profiling de CIMP numa série de 211

pacientes diagnosticados com CCR esporádico.

ADN extraído a partir de 211 CCRs e 43 amostras de mucosa saudável, fixados

em formol e embebidos em parafina, foi convertido pela técnica de bissulfito, e a

metilação dos promotores de cinco genes/loci foi então determinada por qMSP em

tempo real (baseada em SYBR® Green), para avaliar a frequência de CIMP. Foram

então realizadas análises estatísticas para revelar associações com parâmetros clínico-

patológicos, e análises de sobrevivência para avaliar o valor prognóstico de CIMP.

CIMP foi encontrado em 8,5% de todos os casos de CCR e não foi associado

com qualquer dos parâmetros clinicopatológicos e moleculares analisados. Além disso,

CIMP não foi também associado com o prognóstico dos pacientes, tanto no caso da

sobrevivência específica de doença (HR 1,192; CI 95% 0,732-1,941; P = 0,481), como

da sobrevivência livre de doença (HR 0,554; CI 95% 0.241-1,275; P=0,161). No entanto,

metilação aberrante de um dos cinco marcadores que constituem o painel selecionado,


VII

CDKN2A(p16), foi associada com menor sobrevivência específica de doença, mas

apenas em análise univariável (HR 1,578; CI 95% 1,016-2,450; P=0,042).

O fenótipo CIMP não foi associado com a sobrevivência dos pacientes, o que

está de acordo com outros estudos anteriores. Contudo, a técnica de laboratório ou a

sua aplicação com o painel específico selecionado podem não ser adequadas para

avaliar o perfil CIMP, levando a frequências de CIMP mais baixas e à ausência de

associação entre CIMP e qualquer um dos parâmetros testados. Estudos adicionais são

precisos para confirmar estes resultados preliminares.

Palavras-chave: Epigenética, cancro colorectal, lncRNAs, reparação por excisão de

bases, metilação, CIMP, prognóstico.

VIII FCUP Epigenetic Study of Colorectal Cancer: lncRNAs and CIMP Profiling.


IX

TABLE OF CONTENTS

FIGURE INDEX......................................................................................................XI

TABLE INDEX......................................................................................................XIII

LIST OF ABREVIATIONS....................................................................................XV

INTRODUCTION...................................................................................................21COLORECTAL CANCER: GENERAL ASPECTS.....................................................21

Epidemiology and risk factors...................................................................................21Methods of diagnosis...................................................................................................22Histology and molecular etiology.............................................................................22Prognosis and treatment.............................................................................................26

COLORECTAL CANCER EPIGENETICS..................................................................27General aspects, and chromatin and histone modifications.............................27MicroRNAs......................................................................................................................29

LONG NONCODING RNAS & DNA REPAIR.............................................................29LncRNAs involved in colorectal cancer development.........................................29LncRNAs involved in DNA repair..............................................................................32

CpG ISLAND METHYLATOR PHENOTYPE (CIMP) & PROGNOSIS.....................34DNA methylation............................................................................................................34CIMP involvement in colorectal cancer...................................................................34Molecular pathways according to genetic and epigenetic aspects.................37Methods of DNA methylation analysis.....................................................................38DNA methylation as diagnostic biomarker.............................................................38DNA Methylation and CIMP in prognosis and treatment.....................................39

AIMS......................................................................................................................41

PROJECT I............................................................................................................41

PROJECT II...........................................................................................................41

MATERIALS AND METHODS.............................................................................43PROJECT I....................................................................................................................43

Study patients and sample collection......................................................................43Selection of samples and DNA repair assays........................................................43RNA extraction...............................................................................................................44LncRNAs profiling.........................................................................................................44Statistical analysis........................................................................................................45

PROJECT II...................................................................................................................45Study patients and sample collection......................................................................45DNA extraction from paraffinized tissues sections..............................................45Bisulfite conversion......................................................................................................46Primers design and selection.....................................................................................47

X FCUP Epigenetic Study of Colorectal Cancer: lncRNAs and CIMP Profiling.

Quantitative methylation-specific polymerase chain reaction (qMSP)...........48Statistical analysis........................................................................................................49

RESULTS..............................................................................................................50PROJECT I....................................................................................................................51PROJECT II...................................................................................................................54

Patients’ characteristics and CpG island methylation at specific loci............54Prognostic factors for survival: disease-specific survival.................................59Prognostic factors for survival: disease-free survival.........................................64

DISCUSSION........................................................................................................67PROJECT I....................................................................................................................67PROJECT II...................................................................................................................69

REFERENCES......................................................................................................75

APPENDIX I..........................................................................................................81

APPENDIX II.........................................................................................................82

APPENDIX III........................................................................................................83

APPENDIX IV........................................................................................................84

APPENDIX V.........................................................................................................85

APPENDIX VI........................................................................................................86

APPENDIX VII.......................................................................................................87

APPENDIX VIII......................................................................................................87


XI

FIGURE INDEX

Fig.1: Distribution of CRC by anatomical site; illustrative CRC staging, and large intestine wall histological layers................................................................................23

Fig.2: Genetic and epigenetic marks in three proposed pathways to sporadic CRC development................................................................................................................25

Fig.3: Model for DNA repair regulation in CRC by lncRNAs DDSR1, PCAT-1 and HOTAIR.........................................................................................................................33

Fig.4: Estimated distribution of CIN, CIMP and MSI subtypes, and a six-group classification according to MSI and CIMP status in CRC..........................................38

Fig.5: Performance of the classic CIMP panel............................................................56

Fig.6: Comparison between the classic CIMP panel, MINT31 methylation and KRAS mutation status............................................................................................................59

Fig.7: Kaplan-Meier curves analysis for disease-specific survival according to age at diagnosis, AJCC tumour stage, neoadjuvant therapy, CIMP panel and CDKN2A(p16) methylation status...............................................................................63

Fig.8: Kaplan-Meier curves analysis for disease-free survival according to gender, CIMP panel and CDKN2A(p16) methylation status....................................................66

XII FCUP Epigenetic Study of Colorectal Cancer: lncRNAs and CIMP Profiling.


XIII

TABLE INDEX

Table 1: TNM staging system for Colorectal Cancer along with corresponding criteria and anatomic stage (AJCC stage)..................................................................26 Table 2: List of some of the most representative and studied lncRNAs in CRC and associated mechanisms so far described in CRC and other diseases, expression patterns and functions in CRC development............................................................ 31 Table 3: List of primers’ sequences used and respective chromosomal location, size of the generated amplicon, temperature of annealing, GenBank Accession number and specific location in the accessed sequence.........................................50 Table 4: Long noncoding RNAs differentially expressed between the four groups of samples formed HH, HL, TH and TL, before Holm-Šídák correction....................52 Table 5: Long noncoding RNAs differentially expressed between Healthy mucosa and Tumour samples, and samples with Lower and High BER repair capacity, before Holm-Šídák correction.....................................................................................53 Table 6: P-values for the differential expression of long noncoding RNAs between the four groups of samples formed HH, HL, TH and TL, and between Healthy mucosa and Tumour samples or samples with Lower and High BER repair capacity, after Holm-Šídák correction........................................................................53 Table 7: Distribution of clinicopathological and molecular variables for all CRC patients and association with CIMP status................................................................57 Table 8: Association between clinicopathological and molecular variables and each of the five genes/loci constituting the classic CIMP panel...............................58 Table 9: Univariable and multivariable prognostic analyses: disease-specific survival analysis of CRC patients according to represented variables and CIMP panel/markers methylation.........................................................................................61 Table 10: Univariable prognostic analyses: disease-free survival analysis for CRC patients according to represented variables and CIMP panel/markers methylation..................................................................................................................64

XIV FCUP Epigenetic Study of Colorectal Cancer: lncRNAs and CIMP Profiling.


XV

LIST OF ABREVIATIONS 17p – Short Arm of Chromosome 17 18q – Long Arm of Chromosome 18 3’UTR – 3’ Untranslated Regions 5-FU – 5-Fluorouracil 5-mC – 5’-Methylcytosine A – Adenosine ACTB – Beta Actin ACVR2A/1B – Activin A Receptor Type 2A/1B ADP – Adenosine Diphosphate AJCC – American Joint Committee on Cancer AKT – Protein Kinase B ALX4 – Homeobox Protein Aristaless-Like 4 ANRIL – Antisense NcRNA in the INK4 Locus Anti-NOS2A – Anti Nitric Oxide Synthase 2A APC – Adenomatous Polyposis Coli APEX1 – Apurinic/Apyrimidinic Andodeoxyribonuclease 1 ARID1A – AT-Rich Interaction Domain 1A ATM – Ataxia Telangiectasia Mutated ATP – Adenosine Triphosphate AXIN2 – Axis Inhibition Protein 2 BACE1AS – BACE1 Antisense BAX – BCL2-Associated X Protein BER – Base Excision Repair BMP3 – Bone Morphogenetic Protein 3 BOKAS – Natural Antisense Transcript of Bok BRAF – Serine/Threonine-Protein Kinase B-Raf (V-Raf Murine Sarcoma Viral Oncogene Homolog B1) BRCA1/2 – Breast Cancer 1/2 BRG1 – Brahma-Related Gene-1 C – Cytosine c-MYC – Myc Proto-Oncogene Protein (V-Myc Myelocytomatosis Viral Oncogene Homolog) CACNA1G – Calcium Voltage-Gated Channel Subunit Alpha1 G CAP – College of American Pathologists; CapeOX – Capecitabine plus Oxaliplatin CBR/p300 – CREB Binding Protein/EP300 CCAT1-L – CRC-Associated Transcript 1, the Long Isoform CCAT1/2 – CRC-Associated Transcript 1/2 CCE – Colon Capsule Endoscopy CD119 – Cluster of Differentiation 109 CDH1 – Cadherin 1 (E-cadherin) CDK4/6 – Cyclin-Dependent Kinase 4/6 CDKN1A – Cyclin-Dependent Kinase Inhibitor 1A/P21 CDKN1B – Cyclin-Dependent Kinase Inhibitor 1B/P27 CDKN2A – Cyclin-Dependent Kinase Inhibitor 2a/P16 or P14 cDNA – Complementary DNA CDX1 – Caudal Type Homeobox-1 CeRNA – Competing-Endogenous RNA CI – Confidence Interval

XVI FCUP Epigenetic Study of Colorectal Cancer: lncRNAs and CIMP Profiling.

CIMP – CpG Island Methylator Phenotype CIMP-0 – CpG Island Methylator Phenotype-Negative CIMP-H – CpG Island Methylator Phenotype-High CIMP-L – CpG Island Methylator Phenotype-Low CIMP(–) – CpG Island Methylator Phenotype-Negative CIMP(+) – CpG Island Methylator Phenotype-Positive CIN – Chromosomal Instability COX-2 – Cyclooxygenase-2 CpG – Cytosine-Phosphate-Guanine CRABP1 – Cellular Retinoic Acid-Binding Protein 1 CRC – Colorectal Cancer CREB – Camp Response Element Binding Protein CRNDE – Colorectal Neoplasia Differentially Expressed CT – Chemotherapy CTC – Computed Tomographic Colonography CTCF – CCCTC-Binding Factor CTNNB1 – Catenin Beta 1 DAPK – Death Associated Protein Kinase 1 DCC – Deleted in Colorectal Cancer DDR – DNA Damage Response DDSR1 – DNA Damage-Sensitive RNA 1 DFS – Disease-Free Survival DNA – Deoxyribonucleic Acid DNMTs – DNA Methyltransferases DRC – DNA Repair Capacity DSBs – Double-Strand Breaks DSS – Disease-Specific Survival E2F4 antisense – E2F Transcription Factor 4 Antisense EGFR – Epidermal Growth Factor Receptor ERBB2/3 –Erb-b2 Receptor Tyrosine Kinase 2/3 EVL –Enah/Vasp-like EXO1 – Exonuclease 1 EZH2 – Enhancer of Zeste Homolog 2 FAM123B – APC Membrane Recruitment Protein 1 FAP – Familial Adenomatous Polyposis FBN1 – Fibrillin 1 FBXW7 – FBXW7 F-Box and WD Repeat Domain Containing 7 FDA – Food and Drug Administration FIT – Faecal Immunochemical Test FLNC – Filamin C FOLFIRI – Folinic Acid (Leucovorin) plus Fluorouracil plus Irinotecan FOLFOX – Folinic Acid (Leucovorin) plus Fluorouracil plus Oxaliplatin FOLFOXIRI – Folinic Acid (Leucovorin) plus Fluorouracil and Oxaliplatin plus Irinotecan FS – Flexible Sigmoidoscopy FZD10 – Frizzled Class Receptor 10 G – Guanine G9a – Euchromatic Histone-Lysine N-Methyltransferase 2 (EHMT2) GAS5 – Growth Arrest Specific 5 GATA4/5 – GATA Binding Protein 4/5 gFOBT – guaiac Faecal Occult Blood Test


XVII

GR – Glucocorticoid Receptor GSTP1 – GSTP1 Glutathione S-Transferase pi 1 H2A/2B/3/4 – Histone 2A/2B/3/4 H3KX me2/3 – Di/tri-methylation of Lysines X in Histone H3 HAT – Histone Acetyltransferase HCT116 – Human Colon Cancer Cells HDACs – Histone Deacetylases HDMTs – Histone Demethylases HH – Healthy Mucosa with Higher Levels of BER Repair Capacity HIC1 – Hypermethylated In Cancer 1 HL – Healthy Mucosa with Lower Levels of BER Repair Capacity HLTF – Helicase Like Transcription Factor HMTs – Histone Methyltransferases HNPCC – Hereditary Nonpolyposis Colorectal Cancer hnRNPUL1 – Heterogeneous Nuclear Ribonucleoprotein U-like Protein 1 hOGG1 – Human 8-Oxoguanine DNA N-Glycosylase 1 HOPX – HOP Homeobox HOTAIR – HOX Transcript Antisense RNA HOTAIRM1 – HOX Antisense Intergenic RNA Myeloid 1 HR – Hazard Ratio HR – Homologous Recombination HULC – Highly Upregulated in Liver Cancer IBD – Inflammatory Bowel Disease IDLs – Insertion/Deletion Loops IGF2 – Insulin-Like Growth Factor 2 IGF2AS – Insulin-Like Growth Factor 2 Antisense IGFBP3 – Insulin-Like Growth Factor-Binding Protein 3 IGFR – Insulin-Like Growth Factor 1 Receptor IHC – Immunohistochemistry INK4 – Family of Inhibitors of Cyclin-Dependent Kinase 4 Jpx – JPX Transcript, XIST Activator (Non-Protein Coding) KRAS – Gtpase KRAS (V-Ki-Ras2 Kirsten Rat Sarcoma Viral Oncogene Homolog) LET – Low Expression in Tumour LIG3 – DNA Ligase 3 LincRNA – Long Intergenic Non-coding RNA LINE-1 – Long Interspersed Element-1 LncRNA-DDSR1 – Long non-coding RNA-DNA Damage-Sensitive RNA1 LncRNA-JADE – Long non-coding RNA- Jade Family PHD Finger 1 LncRNAs – Long non-coding RNAs LOH – Loss of Heterozygosity LOI – Loss of Imprinting LSD1 – Lysine-Specific Demethylase 1 LUST – LUCA-15-Specific Transcript M – Methylated MALAT1 – Metastasis-Associated Lung Adenocarcinoma Transcript 1 MAP – MUTYH-Associated Polyposis MAPK – Mitogen-Activated Protein Kinase mascRNA – MALAT1-Associated Small Cytoplasmic RNA MBD4 – Methyl-CpG-binding domain protein 4 MDM2 – Mouse Double Minute 2

XVIII FCUP Epigenetic Study of Colorectal Cancer: lncRNAs and CIMP Profiling.

MEG3 – Maternally-Expressed Gene 3 MEG9 – Maternally Expressed 9 MEK – Map Kinase Kinase MGMT – O6-Methylguanine DNA Methyltransferase MINT– Methylated-in-Tumor miRNA – MicroRNAs MLH1 – MutL Homolog 1 MMR – Mismatch Repair MRC – Magnetic Resonance Colonography mRNA – Messenger RNA MSH2/6 – MutS Protein Homolog 2/6 MSI – Microsatellite Instability MSI-H – Microsatellite Instability-High MSI-L – Microsatellite Instability-Low MSP – Methylation-Specific Polymerase Chain reaction MSS – Microsatellite Stable MTOR – Mechanistic Target of Rapamycin MUTYH – MutY DNA Glycosylase MYLKP1 – Myosin Light Chain Kinase Pseudogene 1 MYOD1 – Myogenic Differentiation 1 NcRNAs – Non-coding RNAs NDRG4 – N-Myc Downstream-Regulated Gene 4 Protein NEIL1 – Nei Endonuclease VIII-Like 1 NER – Nucleotide Excision Repair NEUROG1 – Neurogenin 1 NF-KB – Nuclear Factor Kappa B) NGFR – Nerve Growth Factor Receptor NHEJ – Non-Homologous End-joining NRAS – Neuroblastoma RAS Viral (V-Ras) Oncogene Homolog NSAIDS – Nonsteroidal Anti-Inflammatory Drugs NuRD – Nucleosome Remodelling and Histone Deacetylase ORFs – Open-reading Frames OS – Overall Survival P400 – EP400 E1A Binding Protein P400 PALB2 – Partner and Localizer of BRCA2 PARP – Poly (ADP-ribose) Polymerase PBMCs – Peripheral Blood Mononuclear Cells PCAT-1 – Prostate Cancer-Associated Transcript 1 PI3K – Phosphoinositide 3-kinase PIK3CA – Phosphatidylinositol-4,5-Bisphosphate 3-Kinase, Catalytic Subunit Alpha PMS2 – PMS1 Homolog 2, Mismatch Repair System Component PRCs – Polycomb Repressive Complexes PTBP2 – Polypyrimidine Tract Binding Protein 2 PTEN – Phosphatase and Tensin Homolog PTENP1 – Phosphatase and Tensin Homolog Pseudogene 1 PTGS2 – Prostaglandin-Endoperoxide Synthase 2 PVT1 – Plasmacytoma Variant Translocation 1 qMSP – Quantitative Methylation-Specific Polymerase Chain reaction qPCR – Quantitative Polymerase Chain Reaction RAD51 – RAD51 Recombinase


XIX

RAP80 – Receptor-Associated Protein 80 RASSF1A – Ras Association Domain-Containing Protein 1Isoform A/B RET – Ret Proto-Oncogene RFS – Recurrence-free Survival RISC – RNA-Induced Silencing Complex RNA – Ribonucleic Acid RNCR3 – Retinal Noncoding RNA 3 RT – Radiotherapy RTK – Receptor Tyrosine Kinase RUNX3 – Runt Related Transcription Factor 3 SAF – LncRNA Fas-Antisense 1 (Fas-AS1) SEPT9 – Septin 9 SETD2 – SET Domain Containing 2 SFPQ – Splicing Factor Proline/Glutamine-Rich SFRP1 – Secreted Frizzled-Related Protein 1/2 SIRT1 – Sirtuin 1 SLIT2 – Slit Guidance Ligand 2 Smad – Mothers Against Decapentaplegic SMARCC2 – SWI/SNF Related, Matrix Associated, Actin Dependent Regulator Of Chromatin Subfamily C Member 2 SNHG4 – Small Nucleolar RNA Host Gene 4 snoRNAs – Small Nucleolar RNAs SNPs – Single-Nucleotide Polymorfisms SOCS1 – Suppressor of Cytokine Signaling 1 SOX9 – SRY-Box 9 Transcription Factor ssDNA –Single Strand DNA STAT – Signal Transducer and Activator of Transcription SV2C – Synaptic Vesicle Glycoprotein 2C SWI/SNF – SWItch/Sucrose Non-Fermentable T – Thymine TCF7L2 – Transcription Factor 7-Like 2 TFPI2 – Tissue Factor Pathway Inhibitor 2 TGF-β – Transforming Growth Factor-Beta TGFBR – Transforming Growth Factor-Beta Receptor TH – Tumour with Higher Levels of BER Repair Capacity TL – Tumour with Lower Levels of BER Repair Capacity TMEFF2 – Transmembrane Protein with EGF Like and Two Follistatin Like Domains 2 TODRA – Transcribed in the Opposite Direction of RAD51 TP53 – Tumour Protein P53 TUSC7 – Tumour Suppressor Candidate 7 U – Uracil UM – Unmethylated VEGF – Vascular Endothelial Growth Factor VIM – Vimentin WHO – World Health Organization WIF – WNT Inhibitory Factor WNT – Wingless/Integrated WT – Wild-Type XRCC1 – X-Ray Repair Cross-Complementing Protein 1 Zfas1 – Zinc Finger Antisense 1

XX FCUP Epigenetic Study of Colorectal Cancer: lncRNAs and CIMP Profiling.


21

INTRODUCTION COLORECTAL CANCER: GENERAL ASPECTS Epidemiology and risk factors

Cancer was responsible for 8.2 million deaths in 2012, being the second leading

cause of mortality worldwide, and 14 million new cases diagnosed in 2012. Malignant

neoplasms originating in both rectum and colon are typically joint under the same general

designation, representing the third most frequently diagnosed type of cancer in both

sexes (9.7%) or in men (10.0%), and the second among women (9.2%). After lung, liver

and stomach, it is the major cause of cancer-related deaths (8.5%). Regardless of earlier

detection following improved and wider screening over the past two decades in Europe,

as well as effective treatment options, almost half of all individuals diagnosed with CRC

die as a result of the disease.2–4 Analysing specifically the Portuguese population, CRC

represents the most frequent and the major cause of death by cancer (14.5% and 15.7%,

respectively) (online analysis at globocan.iarc.fr).

Generally, males are slightly more affected than females (incidence and

mortality), and present an average age of earlier onset, which could be attributed to

differences at the hormonal level and environmental risk factors predisposition.5–7

However, contrarily to age, gender is not a relevant clinical feature in the assessment of

CRC predisposition. Indeed, according to the American Cancer Society the likelihood for

individuals under 40 years old to develop CRC is 1:1.212, as opposed to 1:24 for

individuals over 70. Moreover, 90% of all cases diagnosed and 93% of deaths were 50

and older.6,8 Therefore, current recommendations for CRC screening are set to start at

50 years for both women and men. However, contrarily to older individuals, incidence

rates in adults younger than 50 years has been increasing, likely related to modern

acquired unhealthy habits and dietary, such as sedentary life, overload of calories and

animal fat consumption. Perhaps unsurprisingly, Europe and Americas account for more

than half of all CRC cases.2,6,9 Indeed, CRC is considered primarily as a “lifestyle”

disease. Behavioural factors associated with increased risk include mainly a diet high in

red or processed meat, but also obesity (measured by waist size), physical inactivity,

heavy alcohol consumption, long-term smoking, and very low intake of fruits and

vegetables. In its turn, higher blood levels of vitamin D, physical activity, higher intake of

dietary fibber, cereal fibber and whole grains, fruit and vegetables, dairy products, milk,

garlic and calcium, and dietary folate have been proposed to be protective. Although not

recommended for CRC prevention, regular use of nonsteroidal anti-inflammatory drugs

(NSAIDS), postmenopausal hormones, oral contraceptives and oral bisphosphonates

have also been associated with a decreased risk. However, for most of these,

22 FCUP Epigenetic Study of Colorectal Cancer: lncRNAs and CIMP Profiling.

contradictory results published or lack of strong evidence and molecular explanation

limits their applicability.2,6,10 For example, folate deficiency was shown to result in

aberrant DNA methylation, mutations and chromosomal aberrations, but some studies

failed to prove a positive correlation, while some others attributed a negative effect to

folic acid fortification/supplementation.10,11

Methods of diagnosis

An early detection of the lesion is fundamental. A significant proportion of CRC

patients are diagnosed with a regional or metastatic stage of the disease.6 Screening

programmes are increasing and include both invasive and non-invasive tests. guaiac

Faecal Occult Blood Test (gFOBT) and Faecal Immunochemical Test for haemoglobin

(FIT) are based on stool analysis and therefore usually more tolerable, with

demonstrated mortality reduction.3 However, their sensitivity for advanced adenomas

and cancer is low to moderate (worse in the case of gFOBT), and their effectiveness is

highly dependent on positivity cut-off level.3,12 Both techniques have been used as initial

screening followed by colonoscopy to confirm positive cases. Such approach appears to

be more costly-effective than colonoscopy only. Although highly invasive and costly,

colonoscopy is still the preferable option due to a high sensitivity coupled with immediate

polyp resection, ampler test time intervals and better outcomes.6 Other less extensively

explored procedures such as Flexible Sigmoidoscopy (FS), Colon Capsule Endoscopy

(CCE), Computed Tomographic Colonography (CTC) and Magnetic Resonance

Colonography (MRC) are predicted to be important alternatives to colonoscopy

screening. Future research will focus on DNA, RNA and protein biomarkers in blood and

stool based tests with higher sensitivity.3,6

Histology and molecular etiology

Before focusing on CRC prognosis and treatment, it is important to pathologically

characterize CRC and describe its classification and diverse etiology. In fact, CRC is a

heterogeneous disease in terms of clinical behaviour and response to therapy, which

correlates with distinct underlying molecular mechanisms and origin.13 The colon (large

intestine) measures almost 150 cm and consists of 4 segments: cecum and vermiform

appendix, colon (ascending, transverse, and descending portions), rectum, and anus. It

is commonly divided in proximal or right-sided colon (cecum, ascending colon, hepatic

flexure, transverse colon, splenic flexure), distal or left-sided colon (descending colon

and sigmoid colon), and rectum (rectosigmoid junction and rectum) [Fig.1].14

Increasingly, it is being recognized that CRC risk factors, tumour characteristics, and

response to treatment may vary across anatomic subsites, mainly between rectum and


23

the rest of the colon.15 Of mention, proximal colon cancers are most common in females,

older patients with mucinous histology, while distal cancers occur more often among

males and younger individuals presenting predominantly absorptive histology.5,16,17

Histologically, the intestine wall comprises the sequence: mucosa, submucosa,

muscularis (or muscularis propria), subserosa and serosa [Fig.1].14 The great majority

(circa 96%) of CRC are adenocarcinomas. Each one of these tumours may take a period

of 30 to 60 years to initiate, plus 1-5 years to 2 decades to progress from previous benign

lesions, known as polyps (most of which are adenomas). The fastest step is

metastization, which may occur a few years after or almost simultaneously with

completion of malignant transformation of the primary CRC.18,19 Although common, less

than 10% of adenomas transform into adenocarcinomas. These adenomatous structures

arise from glandular cells on the epithelium and can grow through the inner layers of the

intestinal wall eventually invading other regional structures as lymph nodes, blood or

lymph vessels, and ultimately metastasize. Liver is the primary metastatic site, followed

by lung. The extent to which cancer has spread at the time of diagnosis is essential to

define its stage and further select treatment and assess prognosis [see Fig.1, Table

1].6,20,21

Fig.1 - Distribution of CRC by anatomical site; illustrative CRC staging, and large intestine wall histological layers. Approximate frequencies (%) of CRC along colon, rectum and anus.22 First onset picture depicts the five AJCC stages,

with red spheres representing regional lymph nodes invasion and metastasis. (Adapted from National Cancer Institute).

The second onset represents the histological division of the large intestine wall, from the lumen to the peritoneum: mucosa

(including surface epithelium, lamina propria and muscularis mucosae), submucosa, muscularis (propria) (with two

differently directed muscle layers), subserosa and serosa. (Adapted from AJCC – 7th Edition Staging Posters).


About 75% of all CRC cases have no apparent predisposing etiology, whereas

the remaining cases are related to familial/hereditary syndromes and Inflammatory

Bowel Disease (IBD). Familial CRC includes known hereditary forms, such as Familial

Adenomatous Polyposis (FAP), Hereditary Nonpolyposis Colorectal Cancer (HNPCC –

also known as Lynch syndrome), MUTYH-Associated Polyposis (MAP), and the

hamartomatous polyposis syndromes. The majority of familial cases have no clearly

identifiable genetic etiology, but it likely comprises less penetrant variations or Single-

Nucleotide Polymorfisms (SNPs).23,24The approximate percentage of distribution for

each condition is as follows: sporadic, ~75%; familial (not known), ~15%; HNPCC, <5%;

FAP, ~1%; IBD, ~1%; MAP, ~1%; hamartomas, <1%.24 Focusing on sporadic CRC, it

can be divided in hypermutated (16%) and non-hypermutated (84%). Particularly,

tumours from the right/ascending colon are more prone to be hypermethylated and to

display elevated mutation rates.25

Almost 30 years ago, Fearon and Vogelstein described a simplified multistep

model for the formation of adenocarcinomas from normal mucosa. The model was based

on the total accumulation of multiple genetic mutations leading to a selective growth

advantage of those cells, with a minimum number of different mutations required –

affecting mostly cell proliferation or DNA damage response (DDR).26 Adenomatous

polyposis coli (APC) inactivation,which is responsible for FAP and approximately 85%

of sporadic CRC mutations, represents the initiating event in adenoma formation, which

is followed by accumulation of multiple mutations inactivating other tumour suppressors

and activating particular oncogenes. Loss of APC or, rarely, mutational activation of β-

catenin (CTNNB1), leads to an aberrant activation of the Wnt pathway. After this

dysplastic phase, adenoma evolution depends on sequential mutations of KRAS (35-

45%) or BRAF (V600E mostly, 8-12%), causing EGFR signalling activation; SMAD2/4

(10-35%) or TGFBR2, which inactivates TGF-β response; and TP53 (35-55%), with loss

of p53 protective function, culminating with carcinoma development [see upper panel of

Fig.2].25–27 However, other genes were found to be affected in Wnt, RTK-MAPK or PI3K,

and TGF-β pathways in CRC. These include, respectively: AXIN2, FBXW7, ARID1A,

FAM123B, SOX9, TCF7L2 and FZD10; ERBB2/3, IGF2, IGFR, NRAS, MEK, AKT,

PIK3CA, PTEN and MTOR; and TGFBR1/2, ACVR2A/1B and SMAD2/3. Additionally,

MYC, PTGS2 (COX-2), ATM and BAX were also shown to be important during CRC

evolution.25 Mutations in DNA mismatch repair (MMR) genes MLH1, MSH2, and less

significantly PMS2 and MSH6 are responsible for Lynch Syndrome. Moreover,

hypermethylation of MLH1 is involved in ~15% of all sporadic CRC, leading to

microsatellite instability. From the 138 driver genes identified (74 tumour suppressor

genes and 64 oncogenes), only 2-8 are important for the development of a sporadic


25

CRC. The rest accounts for “passenger” alterations arising as aftereffect of the process.

Transcriptional regulation, chromatin modification, STAT, Hedgehog, Notch and cell

cycle/apoptosis complete the known list of pathways mutated by driver genes

defects.19,28 However, further explanation and detailing of such cascades goes beyond

the subject of this dissertation.

Fig.2 - Genetic and epigenetic marks in three proposed pathways to sporadic CRC development. At the top:

Fearon-Vogelstein diagram depicting key genes that are inactivated or activated upon mutation (less representative,

inside parenthesis) and/or LOH (18q for DCC and SMAD2/4, and 17p for TP53); and their corresponding pathways (bold)

found to be altered in CRC progression from normal epithelium to carcinoma, encompassing different phases of adenoma

maturation. More recently, three pathways leading to CRC where proposed: traditional (50-70%), alternative (10-30%)

and serrated (10-20%). Each is associated with a particular group of genetic and epigenetic alterations and polyp histology

(respectively, tubular, villous and serrated).

Amongst CRC classifications, one is commonly accepted and considers the

existence of at least three molecular pathways conducting CRC pathogenesis coupled

with genomic instability. Microsatellite instability (MSI; mostly hypermutated tumours),

chromosomal instability (CIN; mostly non-hypermutated tumours), and CpG island

methylator phenotype (CIMP; within both the hypermutated and non-hypermutated

categories) differently affect tumour progression, metastization and treatment response.

A particular etiology behind each pathway explains mutational status, immune response

and other molecular disparities, contributing to a differential prognosis.19

CIN, the most common pathway, is present in 70-85% of all sporadic CRC, and

was also the first to be characterized. However, no consensual explanation for its origin

has been yet reported. It may partly result from defects in chromosomal segregation,

telomere stability, and DDR. CIN tumours frequently display imbalance in chromosome

number (aneuploidy), sub-chromosomal genomic amplifications, loss of heterozygosity


(LOH), chromosome rearrangements, and base substitutions and deletions. Along with

typical karyotypic abnormalities, a specific pattern of altered genes that drive oncogenic

pathways is observed in CIN.19,29 However, it is not clear whether CIN arises from the

evolution of such mutational status or vice-versa. The overall prevalence of genetic

alterations in CIN follows the initial model, described here before, in line with CIN being

observed in adenomas and increasing in tandem with tumour progression. Although

disruption of APC has been proposed to establish a CIN phenotype, it is still

controversial. Nonetheless, CIN was found to be correlated with most cell pathways

altered in CRC.29

Prognosis and treatment

The 1-, 5- and 10-years relative survival rates for CRC are 83%, 65% and 58%,

respectively. When detected at a localized stage, it is highly curable, with 5-year survival

of 90%, in contrast with 70% when spread regionally or 13% in a metastatic stage.7

Colon cancer treatment is greatly dependent on tumour stage. For stages 0 to III,

and some cases of stage IV or recurrence, the primary approach is wide surgical

resection of the lesion, including local excision or polypectomy. Adjuvant chemotherapy

(CT) or radiotherapy (RT) are typically administrated to recurrence cases, stage III-IV

patients, and stage II patients presenting any clinical high-risk features. CT includes

several options, selected according to various factors such as tumour stage and clinical

history/condition of the patient. Generally, 5-FU (5-Fluorouracil) is the basic approach,

to which Leucovorin or a cytotoxic agent (often Irinotecan, Capecitabine or Oxaliplatin)

are coadministered, potentiating 5-FU activity or treatment efficacy.30 Thus, available

treatments include: FOLFOX (Leucovorin, 5-FU, and Oxaliplatin), FOLFIRI (Leucovorin,

5-FU, and Irinotecan), CapeOX (Capecitabine and Oxaliplatin) and FOLFOXIRI

(FOLFOX plus Irinotecan). Biological agents targeting VEGF (Bevacizumab, Ziv-

aflibercept, or Ramucirumab) or EGFR (Cetuximab or Panitumumab) are usually added

to one of the previous therapies, ameliorating the outcome. EGFR inhibitors are only

applicable in tumours without KRAS mutations. Both RT and/or CT can also be used as

neoadjuvant therapy when the tumour is difficult/impossible to resect, as it happens in

most stage IV or recurrence cases. Ablation or embolization techniques might also be

an option to treat some metastasis or recurrent liver tumours. Rectal cancers, more

prone to local recurrence, present a somewhat different treatment, in which neoadjuvant

RT/CT is also proposed for most stage III and some stage II cases.31–33

Table 1 - TNM staging system for CRC along with corresponding criteria and anatomical stage (AJCC stage).33,34

Primary tumour (T)


27

COLORECTAL CANCER EPIGENETICS General aspects, and chromatin and histone modifications

Over the last fifteen years, attention has been driven to epigenetic in detriment of

genetic changes. Currently, epigenetics is defined as heritable and possibly reversible

alterations in the phenotypic expression of the genome, modifying gene expression

without affecting DNA sequence, and encompasses: DNA methylation, histone

modifications, chromatin remodelers and noncoding RNAs (ncRNAs).10 Knudson’s “two-

hit hypothesis” initially referred to gene mutations, both germline or somatic, has been

reformulated to include epigenetic changes. However, it still accurately applies to CRC.

Remarkably, it appears that most genes are aberrantly methylated rather than mutated

TX Primary tumour cannot be assessed T0 No evidence of primary tumour Tis Carcinoma in situ: intraepithelial or invasion of lamina propria T1 Tumour invades submucosa T2 Tumour invades muscularis propria

T3 Tumour invades through the muscularis propria into the pericolorectal tissues (Rectal cancer: T3a <1 mm, T3b 1–5 mm, T3c 5–15 mm, T3d 15+ mm)33

T4a Tumour penetrates into the surface of the visceral peritoneum T4b Tumour directly invades or is adherent to other organs or structures

Regional lymph nodes (N) NX Regional lymph nodes cannot be assessed N0 No regional lymph node metastasis N1 Metastasis in one to three regional lymph nodes N1a Metastasis in one regional lymph node N1b Metastasis in two to three regional lymph nodes N1c Tumour satellite deposits in subserosa or in non peritonealised tissues N2 Metastasis in ≥4 regional lymph nodes (a: 4–6, b: ≥7)

Distant metastasis (M) M0 No distant metastasis M1 Distant metastasis M1a Metastasis in one organ/site (for example liver, lung, ovary, nonregional node) M1b Metastasis in more than one organ/site or the peritoneum

Stage grouping 0 Tis N0 M0 I T1-2 N0 M0 IIA T3 N0 M0 IIB T4a N0 M0 IIC T4b N0 M0 IIIA T1-2 (T2) N1/1c (N2a) M0 IIIB T3-4a (T2-3) (T1-2) N1/1c (N2a) (N2b) M0 IIIC T4a (T3-4a) (T4b) N2a (N2b) (N1-2) M0 IVA Any T Any N M1a IVB Any T Any N M1b


in the average colon cancer genome. For many of those genes aberrant methylation is

the only silencing mechanism observed.35,36 Epigenetics was introduced by CH

Waddington in 193937, and its association with CRC was first discovered in 198338. Since

then, it has been recognized that genetic and epigenetic aberrations are both part of a

complex network that predispose to/trigger the development of each other, leading to

CRC development.39,40

Genomic DNA in eukaryotic cells is packed with specific proteins constituting

chromatin. The repeating unit of chromatin is the nucleosome, which is formed by

wrapping a two-turn “superhelix”, ~145–147 bp, of DNA around a histone octamer core

(two copies of each histone H2A, H2B, H3 and H4). Besides histones, many other

proteins integrate and manipulate chromatin structure.41 Chromatin-remodelling

complexes, through ATP consumption, adjust nucleosomal architecture by mobilizing

(insertion/removal) nucleosomes, altering the configuration of nucleosomal DNA and

histone-octamers, and recruiting other auxiliary proteins. Once formed, large scaffolds

regulate many transcription factors.40,42–44 Based on studies with mouse models and cell

lines, some members of the chromatin-remodelling machinery, such as histone

acetyltransferase (HAT) Tip60, ATPase p400 and nucleosome remodelling and histone

deacetylase (NuRD), modulate the functionality of Wnt-cascade. Moreover, SWI/SNF

complex is also commonly altered in CRC by inactivation upon mutation of ARID1A and

SMARCC2, and promotes metastasis upon mutation of BRG1.45

Chromatin state is another important “tuner” of gene expression, existing in a

condensed inactive state (heterochromatin) or in a noncondensed and transcriptionally

active state (euchromatin). Some residues (mainly lysine and arginine) in the amino-

terminal tails of histones, that project from the nucleosome, are prone to certain post

translational modifications, namely acetylation, methylation, phosphorylation,

ubiquitylation, summoylation, ADP ribosylation, deamination and proline isomerization.44

So far, methylation and acetylation are the two most explored and well-known. Di- and

tri-methylation of H3K4 (H3K4me2 or H3K4me3), and acetylation at H3/H4 (H3K9Ac and

H4K9Ac) are associated with an active state, opposingly to histone hypoacetylation and

tri-methylation at H3K9 (H3K9me3) or H3K27 (H3K27me3), which are considered to be

repressive marks. “Histone code” variations mediate silencing of tumour suppressor

genes and activation of oncogenes, occurring after alterations in the expression and

enzymatic activity of HATs and histone methyltransferases (HMTs) or histone

deacetylases (HDACs) and histone demethylases (HDMTs).46 HDAC1–3, 5, and HDAC7

are upregulated in CRC – at early stages of the disease, in the case of HDAC2.47

Together with class III, these class I HDACs are implicated in the downregulation of

tumour suppressor genes such as caudal type homeobox-1 (CDX1), in the Wnt


29

pathway.48 Lysine specific demethylase 1 (LSD1) is a HDMT, which demethylates H3K4

and H3K9 and has been positively correlated with TNM stage, lymph node infiltration

and metastatic disease in CRC patients.49 Moreover, two multimeric polycomb

repressive complexes (PRCs), PRC1 and PRC2, are transcendent epigenetic regulators

that are able to silence genes either independently or synergistically through its histone

methylation capacity, initiating and maintaining H3K27me2/3, respectively. Also, EZH2

(PRC1 component) and BMI1 are frequently overexpressed in CRC. The former predicts

better recurrence-free survival (RFS) in those patients.45,50

MicroRNAs

From the two thirds of transcripts at some point transcribed from the mammalian

genome only <2% codify any protein, the rest representing noncoding RNA molecules

erroneously believed to present no function.51 MicroRNAs (miRNA, miR) are short RNA

molecules (19–25 ribonucleotides) that mediate posttranscriptional gene repression or

mRNA degradation of target mRNAs, while within RNA-induced silencing complex

(RISC).52 miRNAs are the most widely studied class of ncRNAs, and translationally

control over 60% of protein-coding genes. However, expression of ncRNAs is itself

regulated by numerous proteins, DNA methylation and histone modifications, evidencing

a highly complex network of interactions, which are often deregulated in cancer.53 In fact,

many studies found hundreds of differently expressed miRNAs in CRC, and particularly

connected with every important pathway of the multistep conventional CRC

carcinogenesis.54 Both miRNA122a and miR135a/b downregulate expression levels and

activity of APC and MSH2, mediating adenoma formation. let-7 miRNA family, miR-18a,

-96 and -143 regulate expression of KRAS, while miR-21 and miR-126 are associated

with PI3K pathway. Together with miRNAs regulating c-MYC (miR-17, -18a, -19a/b, -

20a, and -92a), they all play significant roles in an early to advanced adenoma

transition.55 Additional altered miRNAs are implied in pre-malignant to malignant

transformation (p53 regulators miR-16, -143 and -145, and downstream target miR-34a)

or invasion/metastatic phenotype (miR-21, -625, -200 and -126). Moreover,

downregulation of miR-378 and upregulation of miR-127-3p, -92a and -486-3p are

associated with KRAS mutations, while upregulation of miR-31 is instead associated with

BRAF mutations.46,54,55

LONG NONCODING RNAS & DNA REPAIR LncRNAs involved in colorectal cancer development

LncRNAs are simply defined as a class of ncRNAs transcripts longer than 200 nt

and the most representative group among those, usually with no significant open-reading


frames (ORFs) in its sequence.56,57 These poorly conserved RNA molecules present

tissue/cell, disease and spatiotemporal specificity, which supports their superior

applicability as potential biomarkers and treatment-target molecules.57,58 Due to their

inherent proneness to mutations, and hence structural diversity, fast evolutionary

changes come easier to lncRNAs. In line with this, it is not surprising the existence of so

many different functions and classifications attributed to these molecules. They are now

thought to rival the impact of coding-transcripts, being involved in the regulation of most

cellular mechanisms.59

lncRNAs classification is diverse and an ever-changing task. The most varied

classification is based on their function. Genes, proteins, mRNAs, microRNAs are all

targets of regulation by lncRNAs. By interacting with specific proteins, lncRNAs can

either repress, activate, recruit or serve as a scaffold for the assemble of protein

complexes involved in transcription. A common way for lncRNAs to control transcription

is through chromatin-based gene regulation.60 Indeed, several lncRNAs have been

shown to interact with histone modifiers and chromatin remodelling complexes histone

methyltranferases, such as PRCs and G9a protein.61 Although the main regulatory

effects of lncRNAs occur in a pre-translational manner, they are capable of regulating all

processes from gene to protein by different mechanisms. Additionally, lncRNAs are able

to originate miRNAs and snoRNAs, act as molecular decoys or compete for common

binding sites.62

Overall, and particularly in CRC, lncRNAs have successfully helped to clarify

previously unexplainable questions.63,64 Many arrays spotted a differential expression of

numerous lncRNAs between normal and transformed mucosa; according to CRC

development, invasion and metastatic stage63,65; and also in response to treatment, such

as 5-FU66 and radiation67. Additionally, profiles include p53-related68 and MYC repressed

transcripts69, as weel as hypermethylation of genes coding for lncRNAs70. CRC-

associated transcript 1(CCAT1) is upregulated in pre-malignant conditions and all

disease stages in CRC, but not in normal tissues. Therefore, this MYC-regulated lncRNA

has potential to be used for CRC screening, diagnosis, staging and development of novel

therapies.71,72 Another CRC-associated lncRNA, from the same family, is CCAT2, which

is also upregulated only in CRC – involved in cancer progression by promoting its

invasion and metastasis. Also, CCAT2 is correlated with microsatellite stable cancers,

higher expression levels of MYC and potentiation of Wnt signalling pathway.65,73 CCAT1-

L (CCAT1, the long isoform) upregulation in CRC mediates chromatin looping between

the MYC promoter and its enhancers in coordination with CCCTC-binding factor

(CTCF).74 Colorectal neoplasia differentially expressed (CRNDE) is detected in early

adenomas but not in normal mucosa, fostering cell proliferation, migration and invasion.


31

CRNDE promotes Warbug effect and is upregulated in plasma of CRC patients, possibly

being highly valuable for an early diagnosis.63,75 However, many additional lncRNAs are

non-specific to CRC and they include transcripts differently expressed also in other

malignancies.65 As the number of lncRNAs altered in CRC keeps increasing, the

comprehension of the related molecular mechanisms is not parallelly evolving, and only

a small group of transcripts is currently better studied.64 Little of them are implied in early

detection of CRC and fewer in its risk assessment. Nonetheless, lncRNAs often play an

important role in CRC progression, mainly through local invasion and distant metastasis,

which renders them important prognostic biomarkers and treatment options.65,72 Their

underlying mechanisms, expression patterns and functions are described in Table 2.

Table 2 - List of some of the most representative and studied lncRNAs in CRC and associated mechanisms so far described in CRC and other diseases, expression patterns and functions in CRC development.

lncRNA Mechanism Expression Function Ref.

H19 Act as ceRNA for miR-138 and miR-200a; precursor of the RB-inhibitor miR-675. Up or LOI Progression 76–78

HOTAIR Recruitment of PCR2 and LSD1 complexes to HOXD,

silencing HOXD. By supressing SETD2, impairs mismatch repair pathway.

Up Progression Metastasis

65,79

MALAT1 Binds to SFPQ and releases PTBP2; involved in RNA

splicing and small RNA production; promotes cell migration, invasion, and metastasis.

Up Early

Diagnosis Progression Metastasis

63,80,81

HULC Binds to miR-372, and mediates cell invasion and metastasis to the liver. Up Progression

Metastasis 72,82

PVT1 Downregulates Caspase3 and Smad4. Up Progression 83

MYLKP1 Binds MYLK, increasing cell proliferation. Up Progression 84

PCAT-1 By supressing BRCA2, impairs homologous recombination and, therefore, DNA repair. Up Progression 85,86

MEG3 By supressing MDM2, promotes P53 expression, and inhibits tumour growth. Down Progression 72,87

LET Regulates hypoxia signalling Down Progression 88

TUSC7 Association with P53 and inhibition of miR-211, inhibiting tumour growth. Down Progression

Metastasis 72,89

lincRNA-p21

Activated upon DDR by P53, directing P53 to its targets; increases sensitivity to radiation by targeting

Wnt/β-catenin. Down Progression 65,90

PTENP1 Binds to specific miRNAs and PTEN. Down Progression 91

GAS5 Targets GR, inducing apoptosis. Down Progression 92

BRCA2, breast cancer 2; ceRNA, competing-endogenous RNA; GAS5, growth arrest specific 5; GR, glucocorticoid receptor; HOTAIR, HOX transcript antisense RNA; HOXD, Homeobox D cluster; HULC, highly upregulated in liver cancer; LET, low expression in tumour; lincRNA, long intergenic noncoding RNA; LSD1, lysine‑specific demethylase 1; MALAT1, metastasis-associated lung adenocarcinoma transcript 1; MDM2, mouse double minute 2; MEG3, maternally-expressed gene 3; MYLKP1, myosin light chain kinase pseudogene 1; PCAT-1, prostate cancer-associated transcript 1; PRC2, polycomb repressive complex 2; PTBP2, polypyrimidine tract binding protein 2; PTENP1, phosphatase and tensin homolog pseudogene 1; PVT1, plasmacytoma variant translocation 1; SFPQ, splicing factor proline/glutamine-rich; TUSC7, tumour suppressor candidate 7.


LncRNAs involved in DNA repair

Tens of thousands of DNA lesions that each cell experiences per day would

immediately lead to its death if no mechanism of repair was present. DNA damage

response (DDR) is a broad term that includes different molecular responses responsible

for DNA integrity maintenance, and includes DNA damage recognition, recruiting of

mediators, transducers and effectors, culminating in DNA damage repair, activation of

cell cycle checkpoints or even apoptosis.93,94 DNA repair is promptly ignited after the

injury, but it is also highly regulated during the whole process, trying to avoid the ultimate

fate: apoptosis. Evolution has been whittling a limited set of mechanisms, each

responsible for repairing one or few specific DNA detrimental alterations.94 Double-strand

breaks (DSBs) are the less frequent and most toxic DNA lesions, being commonly a

consequence of exposure to UV radiation. DSBs repair is a difficult task for the cell and

consists of two different main pathways: non-homologous end-joining (NHEJ) and

homologous recombination (HR).93,95 Despite of so scarce examples, DSBs repair largely

represents the most studied repair mechanism in light of lncRNAs. DSBs repair-related

transcripts include PCAT-185, HOTAIR96, lncRNA-JADE97, DNA damage-sensitive RNA

1 (DDSR1)98, transcribed in the opposite direction of RAD51 (TODRA)99, antisense

ncRNA in the INK4 locus (ANRIL)100, or natural antisense transcript of Bok (BOKAS)101. The number of lncRNAs simultaneously associated with DNA repair and CRC is

even smaller. DDSR1 was found to be upregulated in different cell lines including colon

cancer cell line HCT116, and regulates early to late phases of DSBs repair response,

starting to mediate the sequester of BRCA1-RAP80 complex away from DNA damage

site, favouring HR. Upon induction by ATM-NF-κB, DDSR1 also mediates repression of

p53 targets, and, at a later stage, greater levels of DDSR1 sequester hnRNPUL1.98 Both

PCAT-1 and HOTAIR are upregulated in CRC, but their DNA-repair associated

mechanisms have not been only described specifically for CRC. While PCAT-1 post-

transcriptionally binds to the BRCA2 mRNA 3ʹUTR, supressing HR pathway85, HOTAIR

represses SETD2 by inhibiting the recruitment of the transcriptional machinery to SETD2

promoter, which reduces H3K36 methylation and consequent recruitment of MSH6-

MSH2 protein heterodimer, culminating with impaired MMR [Fig.3].102 This pathway is

responsible for identifying and excising single-base mismatches and insertion/deletion

loops (IDLs), and is intimately connected with CRC.103 A defective MMR response leads

to the accumulation of DNA errors throughout the genome, more frequently in short

sequences of nucleotide repeats, more prone to these errors, called microsatellites. MSI

is responsible for approximately 15-20% of all CRC cases.104 Tumours with high levels

of microsatellite instability (MSI-H)/unstable are defined as having ≥30% instable loci,

through a reference panel of 5 to 10 microsatellite loci, in opposition to tumours with low


33

levels of microsatellite instability (MSI-L) or microsatellite stable (MSS).105 MSI-H

phenotype is characterized by a proximal location, poor differentiation, mucinous

histology, and dense lymphocytic infiltration, compared to the conventional CIN pathway.

Loss of protein expression of 4 MMR genes (MLH1, MSH2, MSH6, and PMS2) is a test

often verified through immunohistochemistry (IHC) in clinical care. Indeed, MSI-H is a

CRC biomarker with prognosis and treatment prediction value. MSI-H, as hypermutated

tumours progress at a faster pace to malignancy, and usually do not respond to 5-FU

treatment. Moreover, over 40% of MSI-H tumours present mutation of BRAF (V600E).

Overall, however, MSI-H tumours present a better long-term prognosis.19,106

Fig.3 - Model for DNA repair regulation in CRC by lncRNAs DDSR1, PCAT-1 and HOTAIR. Left panel: DDSR1

initially mediates the sequester of BRCA1-RAP80 complex away from the DSB site, favouring HR (1). Upon induction by

ATM-NF-κB, DDSR1 also represses p53 targets (2), and at later stages, greater levels of DDSR1 sequester hnRNPUL1

(3). Middle panel: Upon DSBs, the assembly of RAD51 pre-synaptic filament is accomplished by BRCA1-PALB2-BRCA2

complex. However, PCAT-1 interacts with BRCA2 mRNA inhibiting its transcription and subsequent HR repair. PCAT-1

was shown to be repressed by PRC2-mediated epigenetic silencing. Right panel: HOTAIR inhibits transcription and

phosphorylation (activation) of SETD2, reducing H3K36 methylation and consequent recruitment of MSH6-MSH2 protein

heterodimer, impairing MMR.

In contrast with genetic or epigenetic defects in MMR, base or nucleotide excision

repair pathways (BER and NER, respectively) are largely understudied in CRC. Besides

germline inactivation of BER gene MUTYH (responsible for MAP)24, no other noteworthy

pathological mutations have been described for BER or NER.1 However, methyl-CpG-

binding domain protein 4 (MBD4)107, O6-methylguanine DNA methyltransferase

(MGMT)108 and nei endonuclease VIII-like 1 (NEIL1)109 have been recently described as

targets of promoter aberrant methylation in CRC. Furthermore, polymorphisms in many

BER genes (APEX1,XRCC1, PARP, LIG3, hOGG1, and EXO1) have been linked to

CRC risk.110,111 BER is the main pathway repairing spontaneous, alkylating, and oxidative

small non-helix-distorting chemical lesions of DNA bases112, while bulkier helix-distorting


and more complex lesions, such as pyrimidine dimers and intra-strand crosslinks, are

corrected by NER.113 BER also removes uracil (or its analogues) misincorporated into

the DNA as a result of 5-FU, further linking DNA repair and CRC sensitivity to

treatment.114 Through a functional analysis of the overall DNA repair capacity (DRC) for

BER and NER in a subset of CRC patients, as well as genetic and epigenetic aspects,

Slyskova et al (2012) found no meaningful alterations, indicating that excision repair is

not a major driving factor in malignant transformation, which is consistent with previous

studies.1

CpG ISLAND METHYLATOR PHENOTYPE (CIMP) & PROGNOSIS

DNA methylation

DNA methylation represents the most studied epigenetic area in CRC.115 In fact,

the first epigenetic alteration reported in cancer was global loss of DNA methylation,

represented by 5’-methylcytosine (5-mC), in CRC, and affecting mostly repetitive

transposable sequences, such as LINE-1 and Alu elements.36,38 This is an age-

dependent and early event in CRC development, predisposing to genomic instability,

including loss of imprinting (LOI) and CIN. Accordingly, LINE-1 hypomethylation

inversely associates with MSI and/or CIMP.115 DNA methylation occurs at cytosine bases

preceding guanines, called CpG dinucleotides (C-phosphodiester-G bond), most of

which are methylated in a healthy state. However, there are also unmethylated CpG rich

sequences, called CpG islands, and generally located in the 5’ region of approximately

half of all human gene promoters. CpG islands are 200-2000 bps long, with a CG content

>50% and a ratio of observed to expected CpGs >60%, and are involved in the regulation

of gene expression.54,115,116 When methylated they may induce chromatin conformational

changes, through MBD proteins, hindering promoter assessment and repressing

transcription. In CRC, both hypermethylation and hypomethylation abnormalities are

present, but in a reversed pattern from normal mucosa.46

The addition of a methyl group (-CH3) to a cytosine is catalysed by DNA

methyltransferases (DNMTs) using S-adenosylmethione as the methyl donor compound,

in either a de novo (DNMT3A and DNMT3B) or maintenance fashion (DNMT1).36,46 In

CRC, both DNMT1 and DNMT3B were shown to contribute to CpG methylation and

aberrant gene silencing. Moreover, mutations in DNMT1 and SNPs in DNMT3B have

also been linked with CRC risk.10,39

CIMP involvement in colorectal cancer

It has been increasingly recognized that a distinct methylation pattern appears as

a “function of age”, the so-called “epigenetic drift”, which also affects methylation of


35

promoters.115 Throughout the rest of this text the term “methylation” will be applied in the

sense of gene promoter hypermethylation unless otherwise stated. DNA methylation in

both normal-appearing mucosa and CRCs (age-related methylation, type A methylated

genes) may precede tumour formation, arising in close relation with epigenetic

microenvironment and external factors, whilst DNA methylation specifically in CRCs

(cancer-specific methylation, type C methylated genes) seems to be a less random

process, and is associated with a more limited number of genes and a subset of CRCs

– which then evolve along a CIMP pathway.117 The molecular causes underlying such

methylation are not well-understood, but there are multiple models for cancer-related

aberrant methylation, encompassing mechanisms such as overexpressed, hyperactive,

or misdirected DNMTs, dysregulation of associated ncRNAs, unrepaired halogenated

DNA damage products mimicking 5-mC, and impaired barrier elements.36

Hundreds to thousands of genes are aberrantly methylated in the average CRC

genome, and although no sharp distinction between type A and type C genes has been

made, many of the CRC-specific hypermethylation events have been linked to the same

important pathways targeted by mutational events.36,39,54 The term “CpG island

methylator phenotype,” or CIMP, was first coined in 1999 by Toyota, with Baylin, Issa

and others to characterize some tumours presenting a distinct phenotype of

simultaneous and intense promoter hypermethylation of some tumour suppressor genes,

leading to progressive genetic silencing and tumourigenesis, even in the absence of any

genetic mutations. According to the same study, CRCs can be divided in CIMP− (CIMP-

negative) or CIMP+ (CIMP-positive), respectively displaying rare methylation or

simultaneous aberrant methylation of several genes.117 One of the first and best studied

alterations was biallelic promoter CpG island methylation of MLH1, which unveiled a

strong link between CIMP and MSI-H tumours.39 Indeed, CIMP+ tumours have been not

only associated with MSI-H phenotype, but also with older age, female sex, mucinous

cell differentiation, smoking, BRAF and less often KRAS mutations.39 Approximately 20%

of CRCs are CIMP tumours36,118, rarely occurring in rectal cancer and increasing linearly

up to the ascending colon.119

To define CIMP, promoter methylation of a panel of specific genes is evaluated,

with some of them being more valuable than others. However, which specific methylated

loci should be used to describe CIMP is not standardized.54,120 The so-called “classic”

panel of Park et al., later described by Issa117, comprises CpG islands in MLH1,

CDKN2A(p16), and methylated in tumours (MINTs) 1, 2, and 31 loci, and provides a

simplified and representative approach to define CIMP. The five methylation markers

have distinct functions. MINT markers correspond to the promoters of unique genes

except MINT2; MINT1 corresponds to synaptic vesicle glycoprotein 2C gene (SV2C),


and MINT31 corresponds to a CpG island upstream of the calcium channel CACNA1G

gene.121,122 P16 is an inhibitor of cyclin-dependent kinase 4 (CDK4) and CDK6, which is

associated with aging and functions as a tumour suppressor, leading to unrestrained cell

proliferation upon genetic or epigenetic inactivation.123 In Park’s work, the selected

technique to analyse methylation was methylation-specific PCR (MSP).124 These 5

genes/loci were tested first in 1999 by Toyota et al. along with other 25 newly cloned

differentially methylated DNA sequences, and were later selected based on their

frequent methylation.117,124 In 2006, using MethyLight technology, Weisenberger et al.

proposed a new robust 5-gene panel (CACNA1G, IGF2, NEUROG1, RUNX3, and

SOCS1), further supporting CIMP as a distinct molecular trait of CRC.125 Nevertheless,

it does not seem to outperform the “classic” one.116 Both studies classified tumours as

CIMP+ when more than 1 marker was methylated. The dichotomized CIMP classification

adopted in those articles, while being the first defined and the most common, is not the

most informative. This bimodal distribution failed to explain CIMP+/MSS tumours for

example, which were then shown to be better clarified in a tri-modal partition of CIMP-

High (H), CIMP-Low (L) and CIMP-0.116,118,126,127 In fact, Ogino et al. quantified DNA

methylation (MethyLight) also in 5 CIMP-specific gene promoters (CACNA1G,

CDKN2A(p16), CRABP1, MLH1, and NEUROG1), defining tumours presenting 4-5/5

methylated markers as CIMP-H, 1-3/5 methylated markers as CIMP-L, and 0/5

methylated markers as CIMP-0 tumours.127 Using another large cohort, the same author

tested the prior markers plus IGF2, RUNX3, and SOCS1 to classify CRC as CIMP-H

when 6-8/8 markers were methylated, CIMP-L when only 1-5/8 were methylated, and

CIMP-0 if no promoter was found to be methylated.128 Moreover, both “classic” and “new”

panels have also been applied in a tri-model classification, and they may be further

developed to contain additional loci.118

The characteristics of the three CIMP groups are not well defined, but they do

present independent associated features. Regardless of MSI status, CIMP-H tumours

correlate with proximal tumour location, serrated pathway, older age, female gender,

poor differentiation,signet ring cells, high BRAF and low TP53 mutation rates, loss of

nuclear p27 (CDKN1B), LINE-1 methylation, inactive CTNNB1 and PTGS2, and

expression of DNMT3B, p21 (CDKN1A), TGFBR2 and SIRT1.11,120 CIMP-H represents

~15-20% of all CIMP tumours, and although it is more related to MSI-H, both MSI-H

CIMP-H and MSI-L/MSS CIMP-H tumours exist.120 CIMP-L tumours are usually MSS or

MSI-L, characterized by CIN, and associated with male gender and KRAS mutations. In

fact, KRAS and BRAF mutations are mutually exclusive and seem to play an important,

yet still unclarified, role in CIMP development.46 By its turn, CIMP-0 tumours are

associated with CIN, wild-type KRAS/BRAF, distal colon and show no sex predilection.120


37

Furthermore, CIMP-negative tumours have been occasionally split into two different

subtypes, one associated with TP53 mutations and distal location, and the other one

showing a low frequency of hypermethylation or cancer-specific gene mutation, while

mostly located in the rectum.46

Distinctive molecular subclasses of MSI/CIMP tumours have been proposed

when classifying CRCs. However, because of subtle differences between CIMP-L and

CIMP-0 or MSI-L and MSS, only 6 groups are comfortably distinguished. The main

disparities described between MSI-H CIMP-H (10%) and MSI-L/MSS CIMP-H (5-10%)

tumours are MLH1 promoter methylation (MSI-H CIMP-H) and their respective

association with good or poor prognosis. MSI-H CIMP-L/0 (5%) includes mainly Lynch

syndrome, but also sporadic CRC. Unlike MSI-L CIMP-L tumours (~5%), this subtype is

preferably located in proximal colon and is not correlated with MGMT methylation and

loss. In addition, MGMT methylation is also the main difference between MSI-L CIMP-L

and MSS CIMP-L (30-35%) tumours. The characteristics of the remaining and biggest

group, MSI-L/MSS CIMP-0 (40%), greatly overlap with those described above in light of

the first CIMP-negative group [Fig.4].120

Molecular pathways according to genetic and epigenetic aspects

The statement that epigenetic changes take place at early stages of adenoma

formation inspired the division of sporadic CRC formation into three pathways, firstly by

Issa and later by Coppedè.46,129 The majority of sporadic CRCs originate in conventional

villous and/or tubular adenomas, following the classic adenoma-carcinoma sequence130,

and are further divided according to its association with KRAS (alternative pathway) or

APC (most traditional pathway) mutations and CIMP-L or CIMP-0, respectively. Common

features between them include CIN, MSI-L/MSS status and TP53 mutations (specifically

in the distal colon, for the KRAS-mutated pathway).46 Moreover, while the APC-mutated

pathway is the most typical (50-70%), the KRAS-mutated pathway (10-30%) is correlated

with poor prognosis and unresponsiveness to 5-FU and Cetuximab.46,129 More recently,

another new “alternative” to the conventional adenoma-CRC pathway with unique

features has been described (10-20%), involving instead serrated polyps as the

precursor lesion and evolving through suppressive methylation of many key genes. This

is the route through which many CIMP tumours arise, and is also associated with BRAF

and KRAS mutations (but not APC or CTNNB1), proximal location, and MSI

[Fig.2].46,115,129 Importantly, CIN, MSI and CIMP are not mutually exclusive. Indeed, up to

25% of MSI and 33% of CIMP+ tumours can exhibit chromosomal abnormalities, while

most MSI/CIN– CRCs are also CIMP+, and up to 12% of CIN+ tumours are MSI-H

[Fig.4].131–133


Fig.4 - Estimated distribution of CIN, CIMP and MSI subtypes, and a six-group classification according to MSI and CIMP status in CRC. CIN is the most common subtype in CRCs (circle shape, 70-85%), followed by MSI (dashed triangle,

~15-20%) and CIMP (dotted triangle, ~20%). However, these frequencies are still a controversial topic. Some CIMP

tumours (mostly CIMP-L) also display a CIN phenotype, while most CIMP-H tumours are also microsatellite instable.

Moreover, a reduced number of tumours may present both chromosomal and microsatellite instability. Ogino and Goel

(2008) also proposed a classification of CRC according to MSI/CIMP status into six groups, albeit three of them are not

well-defined. (Right panel was adapted120).

Methods of DNA methylation analysis The heterogeneity in CIMP-related studies goes far beyond the panel and the

threshold selected. Besides the different clinical characteristics of the population

(including clinical stage, treatment and location of the tumour), specimen preservation

(either cryopreservated or formalin fixed paraffin embedded) and laboratory methods to

assess gene methylation greatly varies between studies. MSP and MethyLight are the

two most preferred techniques, followed by bisulfite pyrosequencing and combined

bisulfite restriction analysis (COBRA).118 MSP is a rapid and cost-effective qualitative

method of analysis that uses bisulfite-modified DNA as a template for PCR amplification

with two primer sets – specific for methylated (MSP) and unmethylated (classical PCR)

sequences. Quantitative variations of this technique based on real-time PCR include

detection through an intercalated dye like SYBR® Green or by a TaqMan® probe

(MethyLight). These high-throughput, specific and sensitive assays determine the level

of methylation upon normalization of the signal usually to an Alu- or β-actin-based control

reaction.134,135

DNA methylation as diagnostic biomarker (Biological) biomarkers have been classically defined as “a characteristic that is

objectively measured and evaluated as an indicator of normal biological and pathological

processes, or pharmacologic responses to a therapeutic intervention”.136 An ever-

increasing number of studies have demonstrated the potential for using methylated DNA

as biomarker for the early detection of CRC and, less representatively, its application as

a prognostic or predictive biomarker.137 Robust and reliable non-invasive biomarker


39

assays for the detection of early CRC are needed, as currently it represents the most

effective strategy to reduce mortality. In fact, somatic mutations are relatively rare

compared to DNA methylation alterations in the early stages of CRC tumourigenesis.115

Both blood- or stool-based biomarkers have been proposed. Commercially available

tests include FDA approved analysis of Vimentin (VIM) gene methylation (ColoSure™)

and plasma-based test of aberrantly methylated Septin 9 gene (SEPT9), in Europe

(EpiproColon® 1.0, ColoVantage® and RealTime mS9).115 Other methylated biomarkers

in “circulating DNA” (ALX4, DAPK, NGFR, HPP1, NEUROG1, RUNX3 and TMEFF2),

stool (ATM, BMP3, FBN1, GATA4/5, GSTP1, NDRG4, SFRP1 and TFPI2) or in both

(APC, CDKN2A(p16), HLTF, MLH1, MGMT, RASSF2A, SFRP2 and WIF) were

proposed for CRC screening.115,137,138 Moreover, methylated BMP3, methylated NDRG4,

and mutant KRAS combined test (Cologuard®) was also FDA approved as a stool DNA-

based assay, and showed a greater overall sensitivity than FIT test for CRC or early

adenoma detection.139 Notably, multitarget DNA tests and/or combination with

conventional approaches are likely to improve the sensitivity to detect the lesion.140

DNA Methylation and CIMP in prognosis and treatment Clinical decisions following prognosis in CRC are currently based on tumour

staging and histopathologic characteristics – categories I and II of prognostic factors,

respectively, according to the College of American Pathologists (CAP).115,141 However,

such approach is fallible, as illustrated by numerous patients with the same stage, which

progress differently, surviving shorter or longer periods. Selection of specific methylated

DNA signatures seems to be highly feasible for the development of prognostic

markers.115 Methylation of APC, CDKN2A(p14) or RASSF1A was associated with poor

prognosis in a subset of patients independently of tumour stage or differentiation142, while

HOPX and RET were correlated with worse prognosis of stage II and III CRC,

respectively. Other genes whose aberrant methylation has been associated with poor

prognosis include CDKN2A(p16), IGF2 and extracellular matrix remodelling pathway-

associated genes (IGFBP3, EVL, CD109 and FLNC).46 Methylation of HLTF and

TMEFF2 in serum was independently associated with poor outcome.143 In opposition,

methylation of MGMT or MLH1 was linked to a more favourable prognosis.142,144

Moreover, methylation of genes targeted by the polycomb group of proteins (SFRP1,

MYOD1, HIC1 and SLIT2) was also associated with good prognosis in CIMP– male

patients.115 Because different pathways are commonly affected in CRC, selecting a panel

of different biomarkers will potentiate the accuracy of the test. Therefore, among all

biomarker candidates, CIMP status is by far the most promising indicator for

prognosticating CRC patients in terms of phenotypic presentation, therapeutic response


and survival outcomes.115,145 CIMP+ tumours have been independently associated with

shorter survival in many studies, irrespective of MSI status, particularly in patients with

early and locally advanced CRC. Moreover, CIMP status was also shown to be a

negative prognostic factor in patients with metastatic colorectal cancer treated with CT.146

However, conflicting results have been reported as well, with some studies describing a

null association between CIMP-H and CRC prognosis, or even noticing a better

prognosis after CT.118,147 Since most of CIMP-H-related clinicopathological and

molecular features overlap with those for MSI cancers, CIMP status is believed to

influence the good prognosis of CRC that is attributed to MSI. Therefore, evaluation of

both CIMP and MSI is highly recommended when stablishing prognosis.145 Additionally,

such analysis has been proved to depend also on the specific location of the tumour.145

Bae et al. found that CIMP+ tumours correlated with shorter disease-free survival (DFS)

and overall survival (OS) in distally but not in proximally located tumours, while MSI was

correlated with better survival only in proximal tumours.148 Given that most CIMP+

tumours present BRAF mutations, then CIMP biomarker is also predictive for mutational

profile in CRCs.125 In fact, BRAF mutated CRCs may contribute to shorter survival time

in CIMP+ MSS tumours.149

Although plenty of studies have been evaluating the predictive value of CIMP in

treatment response mainly to 5-FU, no solid conclusion has been reached. Some studies

agreed that adjuvant CT conferred a DFS and OS benefit among CIMP+ stage II and III

CRC patients, while others concluded to the contrary or found no significant predictive

valour.118 Nonetheless, the administration of 5-FU to treat CIMP tumours is not currently

recommended. One interesting study showed the benefits to stage III CIMP+ MSS

patients after the addition of Irinotecan to 5-FU/Leucovorin therapy (FOLFIRI). CIMP was

more strongly associated with a better response to the addition of Irinotecan than MMR

status.150

Besides being an appealing diagnostic, prognostic, and predictive biomarker,

alterations of methylation status are also potential pharmacologic targets, as they are

reversible, stable and early-in-development events. Agents inhibiting DNMTs and

HDACs can be applied to reactivate epigenetically silenced tumour suppressors. DNA

demethylating drugs 5-Azacitidine and 5-Aza-2′deoxycitidine (Decitabine), and HDAC

inhibitors Vorinostat and Valproic acid are currently approved for the treatment of some

malignancies. The combination of both groups of inhibitors has been suggested to be a

better strategy due to a more synergistic effect, as well as its coadministration with

chemotherapeutic drugs. Despite their low specificity and high toxicity, further preclinical

investigations and several clinical trials, in order to establish the applicability of these

and other related agents in CRC treatment.30,54


41

AIMS

The overall aim of the present dissertation is the epigenetic characterization of

sporadic CRC in light of lncRNAs (Project I) and DNA methylation (Project II),

respectively in two independent population-based sets. Particularly, the first project

encompassing the discovery of new transcripts altered in CRC and associated with DNA

excision repair pathways, and the former focused on the analysis of promoter CpG

islands hypermethylation pattern (CIMP tumours) and further determination of CIMP

prognostic value.

PROJECT I

The objective of this work was primarily to stablish the expression profile of ninety

disease-related lncRNAs in twenty tissue samples, which were equally divided into four

groups according to being either healthy mucosa or CRC lesions, and presenting either

lower or higher DNA repair capacity for BER DNA repair pathway; in order to find a

possible role for lncRNAs in sporadic CRC tumourigenesis in association with BER

functionality, and ultimately finding new biomarkers or treatment-targets.

PROJECT II

The main goal of this second project was the profiling of CIMP status in tissue

samples from a subset of 211 CRC patients and 43 controls, by measuring the promoter

methylation of the “classic panel” of five genes/loci, through real-time qMSP (SYBR®

Green-based) with bisulfite converted DNA; and further study the possible association

with other molecular and clinicopathological features and patients’ prognosis.


43

MATERIALS AND METHODS

PROJECT I

Study patients and sample collection The study included twenty tissue specimens isolated from seventeen patients

with sporadic primary CRC who underwent surgical resection, selected from a previous

subset of 70 patients included in Slyskova’s work.1 Patients were recruited between 2009

and 2011 at the Thomayer Hospital (Prague, Czech Republic), the General University

Hospital (Prague, Czech Republic), and Teaching Hospital and Medical School of

Charles University (Pilsen, Czech Republic). All patients signed informed consent. Ethics

approval was granted by the appropriate committees at the 3 hospitals. Tumour tissue

and adjacent healthy colon/rectal tissue (5–10 cm distant from the tumour) were resected

from all patients. All subjects were of the same ethnicity (Caucasian). Tumour and

adjacent normal tissues were deep frozen immediately after extraction and stored at

−80°C.

Selection of samples and DNA repair assays From the 70 paired samples tested by Slyskova and colleagues, 30 samples (24

paired, 3 CRC and 3 from normal mucosa) were selected based on RNA integrity number

(RIN) (≳ 5) measured before, and available expression of BER genes and related DNA

repair capacity data. The samples were then scored according to the values of BER-

DRC, and the median was calculated for each of the two groups of samples (CRC versus

control). The five highest and lowest values were selected, allowing four groups to be

formed, each with five samples, namely: CRC with higher BER-DRC, CRC with lower

BER-DRC, healthy mucosa with higher BER-DRC and healthy mucosa with lower BER-

DRC.

Although the determination of DRC values was not included in the present work,

as it had been already performed, a brief and general explanation of the procedure will

be next attended. Firstly, proteins were extracted from tissueand protein concentration

was measured by a Fluorescamine assay (Sigma-Aldrich Chemie GmbH, Steinheim,

Germany), with a NanoDrop® 3300 (Thermo Scientific, Wilmington, DE, USA). In vitro

repair assays were adopted as previously described150 and implemented using a 12-gel

slide format.105,151 Protein extracts were then incubated with substrate DNA from human

PBMCs treated with Ro 19-8022 (Hoffmann-La Roche, Basel, Switzerland) for 5 min,

and irradiated by a 500 W halogen lamp to induce 8-Oxoguanines, which are known to

be repaired specifically by BER. Levels of DNA strand breaks, generated during removal


of lesions, reflect the repair activity of the extract. After pipetting each extract per agarose

gel and the period of incubation, the protocol followed was the same as described before

for the Comet Assay.153 Each extract was also incubated with DNA from untreated

PBMCs to determine non-specific endonuclease activity of the extract. Finally, slides

were stained with SYBR® Gold (Invitrogen, Carlsbad, CA, EUA), and comets were

scored using a Nikon fluorescence microscope. DRC data were evaluated as tail DNA%

(%T).1

RNA extraction Total RNA was extracted from tissues using AllPrep™ DNA/RNA mini kit (Qiagen,

Hilden, Germany) according to manufacturer’s instruction. Concentration and purity of

all RNA samples were determined spectrophotometrically by measuring their optical

density (A260/280>2.0; A260/230>1.8) using a NanoDrop® ND-2000c (Thermo

Scientific, Wilmington, DE, USA). Additionally, RIN was checked using an Agilent

Bioanalyzer 2100, with a RNA 6000 Nano LabChip® (Agilent Technologies, Palo Alto,

CA, USA), following the protocol provided. The quality of some RNA samples was assed

by electrophoresis using 2.5% Ethidium Bromide (EtBr)-stained agarose gels instead.

LncRNAs profiling The simultaneous expression of 90 lncRNAs, five housekeeping reference

controls and one negative control was determined using the disease-related Human

LncProfiler™ 96-well qPCR Array Kit (cDNA synthesis kit and qPCR array) according to

the instructions of the manufacturers (System Biosciences, Mountain View, CA, USA).

Each kit allows 20 profiles to be performed. Off mention, lncRNA cDNA synthesis

reaction setup includes three different steps, namely polyadenylation, annealing of

adaptor, and conversion to cDNA. The initial step greatly enhances cDNA synthesis

yields of lncRNAs, potentiating the detection by qPCR. 5 µL of total RNA from each

sample (diluted to ~200-400 ng/µL) were cDNA converted and next submitted to real-

time PCR. Briefly, reaction mixtures for each 96-well qPCR plate consisted of 1400 µL

2X SYBR® Green PCR Master Mix (Applied Biosystem, Foster City, CA, USA.), 1400

µL of RNase-free water (QIAGEN GmbH, Hilden, Germany) and 20 µL of cDNA. The kit

provided primers in a plate, which were resuspended with 44 µL of RNase-free water

(QIAGEN GmbH, Hilden, Germany) per well before being use. 2 µL of each primer pair

and 28 µL of reaction mixture were loaded in each well of the qPCR plate. Thermal

cycling conditions consisted of initial incubation at 50°C, a denaturation step of 95˚C for

10 min, followed by 40 cycles of denaturation at 95˚C for 15 sec, and annealing and

extension at 60˚C for 1 min. An additional dissociation stage was included (95°C for 15


45

seconds, 60°C for 15 seconds, followed by a slow ramp to 95°C). Real-time PCR

analysis was performed with an Applied Biosystems® 7500 Real-Time PCR Sequence

Detection System. Finally, the cycle number at which the reaction crossed a threshold

(CT) was determined for each gene. Analysis of the RT-qPCR data was performed using

SDS version 1.3.1 software (Applied Biosystem, Foster City, CA, USA) as previously

described154. Raw CT data for each lncRNA was normalized to the geometric mean of

the five control genes, per plate (ΔCT = CTlncRNA − CTcontrols). The relative expression

levels of target lncRNAs were determined by the equation 2−∆CT. Fold change values

were calculated between two groups of interest: 2-(∆ CTgroup of interest 1 – ∆CTgroup of interest 2).

Statistical analysis Expression data from lncRNAs profiling were statistically evaluated using

GraphPad prism software version 7.0. P-values of less than 0.05 were considered

statistically significant. P-values were adjusted according to Holm-Šídák correction for

multiple comparisons.

PROJECT II

Study patients and sample collection 213 samples from sporadic and primary CRCs were obtained from a wide series

of patients diagnosed and submitted to tumour removal surgery at the Portuguese

Oncology Institute – Porto, Portugal, between November 1994 to March 2012, with no

previous history of CRC. However, almost 92% of all patients studied were diagnosed

between 2005 and 2012. All CRCs were extracted from primary tumours. Tissues were

routinely fixed and paraffin-embedded for standard pathologic examination, allowing for

tumour classification and World Health Organization (WHO)/American Joint Committee

on Cancer (AJCC) grading and staging.155,156 Additionally, an independent set of 50

paraffin-embedded normal colorectal mucosa from patients not diagnosed with CRC or

IBD was used as control. Relevant clinical data were collected from clinical charts [see

Results section – Table 7]. The study was approved by the institutional review board

(CESIPOFG-EPE 120/015).

DNA extraction from paraffinized tissues sections A representative paraffin block from each patient was selected and 12 serial 8-

micrometres thick sections were cut and placed on glass slides, from which two were

H&E stained (initial and final slides). Next, an experienced pathologist delimited the area

of tumour to be macrodissected, in the corresponding H&E stained slides. Other six non-


stained slides were deparaffinised using Xilol and Ethanol 100%, 90%, 70% and 50%,

following initial incubation at 55 ºC for 30-60 min to melt paraffin. A disposable sterile

scalpel blade was then used to macrodissect the selected tumour areas from the slides

with the addition of some drops of digestion buffer (Tris-HCl 1M, EDTA 0.1M, Tween 20

and sterile bi-distilled water (B.Braun, Melsungen, Germany)), by superposition to the

proper H&E stained slide. The removed portions were subsequently placed in labelled

1.5 mL tubes, with 1000 μL of digestion buffer plus Proteinase K (20 mg/mL, 25 μL)

(Zymo Research Corp., Irvine, CA, USA), and left incubating at least overnight at 55ºC,

until total digestion was accomplished. An extra 15 μL volume of Proteinase K was added

to facilitate complete digestion of some samples.

DNA was extracted from tissue according to the standard Phenol-Chloroform

procedure157, using 500 μL of Phenol-Chloroform solution at pH 8 (Sigma-Aldrich and

Merck KGaA, Darmstadt, Germany) in Phase Lock Gel Light tubes (5 Prime, Hamburg,

Germany). After centrifuging the tubes for 15 min at 13 000 rpm, the upper aqueous

phase containing DNA was transferred to a new tube, and then precipitated at -20ºC

overnight using chilled Ethanol 100% (twice the volume of the aqueous phase) (Merck

KGaA, Darmstadt, Germany), Ammonium Acetate 7.5 M (1/3 volume) (Sigma-Aldrich

Chemie GmbH, Steinheim, Germany) and Glycogen (2 μL) (Ambion, Austin, TX, USA).

This step was followed by two centrifugations at 13 000 rpm for 20 min with 70% Ethanol,

and the pellets were then air dried and eluted with bi-distilled water (B.Braun, Melsungen,

Germany). After DNA elution, concentrations were determined using NanoDrop™ Lite

Spectrophotometer (Thermo Scientific, Wilmington, DE, USA).

Two of the total 213 sporadic CRC cases and seven control samples lacked

enough material to be extracted, or re-extracted due to low amounts of DNA yielded.

These were excluded from this work. The next procedures were conducted for the

remaining 211 CRC cases and 43 controls.

Bisulfite conversion First introduced by Frommer et al (1992), bisulfite conversion is the gold-standard

technology for detection of DNA methylation; grounded on the finding that Sodium

Bisulfite treatment of cytosine and 5-methylcytosine has different consequences,

originating different DNA sequences for methylated and unmethylated DNA. In this

regard, cytosines in single-stranded DNA are converted into uracil residues and

recognized as thymine in subsequent PCR amplification and sequencing, while 5mCs

are immune to this conversion and remain as cytosines allowing them to be distinguished

from unmethylated cytosines. The procedure includes initial denaturation of DNA double-

strand, followed by sulfonation of unmethylated cytosines, giving origin to a cytosine


47

sulfonate, then deamination and finally desulfonation, thus losing the bisulfite group and,

finally, becoming uracils. Converted DNA strands are no longer self-complementary,

permitting the evaluation of DNA methylation along the DNA single strand (ssDNA).158,159 The required volume of DNA to achieve the final quantity of 1000 ng of DNA was

diluted in sterile double-distilled water to a total volume of 20 μL in a PCR tube, according

to the specified concentration of each sample. Due to the low concentration of some

samples, the quantity of DNA extracted from those samples was instead adjusted to 500

ng, 600 ng or 750 ng, and equalized at the last step of the conversion procedure.

DNA denaturation and bisulfite conversion were processed into one-step using

the EZ DNA Methylation-Gold™ Kit (Zymo Research Corp., Irvine, CA, USA) according

to manufacturer’s instructions. Briefly, 130 μL of the CT conversion reagent were added

to 20 μL of each DNA sample tube. The samples were then transferred to a Veriti® 96-

Well Thermal Cycler (Applied Biosystems Inc, Foster City, CA, USA) running under the

following steps: 98 ̊C for 10 min, 64 ̊C for 3 h, and storage at 4 ̊C. 600 μL of M-Binding

Buffer were added to a Zymo-Spin IC™ column, followed by the samples, and after 10

min the mixture was centrifuged at 10 000 rpm for 30 sec. Each column was washed

using 100 μL of M-wash buffer, with a new centrifugation. 200 μL of M-Desulphonation

buffer were then added and the plate was left at room temperature during 20 min,

followed by another centrifugation. Two consecutive steps including washing (200 μL of

M-wash Buffer) and centrifugation were performed. Each column was then transferred

to a new tube and 30 μL of bi-distilled water were directly added to the centre of each

column. 5 min later, the samples were centrifuged at 12 000 rpm for 30 sec to elute the

DNA. This step was repeated with an additional 30 μL volume of double-distilled water,

completing a total volume of 60 μL added. For those samples with lower amounts of

DNA, the total elution volumes applied were respectively 30 μL, 36 μL and 45 μL.

Universal Methylated Human DNA Standard (Zymo Research Corp., Irvine, CA, USA)

was used as DNA methylation control, in which case 10 μL were used to prepare the

initial dilution, and a total volume of 20 μL (10 μL + 10 μL) of bi-distilled water was added

to elute DNA.

Primers design and selection A subsequent PCR process is necessary to determine the methylation status of

targeted loci by using specific methylation primers after the bisulfite treatment. Therefore,

new primers specific for methylation were design using Methyl Primer Express®

Software v1.0 (Applied Biosystems Inc. Foster City, CA, USA). After copying the specific

gene/locus sequence from GenBank® (NCBI) to the program, the proper CpG island

was selected, and suggested primer sequences were scored by the program. According


to the characteristics of each primer pair analysed with NetPrimer (Premier BioSoft, Palo

Alto, CA, USA), and the pretended location, one of the proposed pairs was selected per

gene/locus. To design the new primers each gene/locus CpG islands were investigated

first. The studies by Toyota and colleagues (1999)117,160 and Kondo et al (2003)161 were

used as reference to selected the proper (region of) CpG island upstream of each of the

five genes/loci CDKN2A(p16), MLH1, MINT1, MINT2 and MINT31; as these articles

represent pilot reports and the basis for many other studies characterizing the same

genes/gene panel, regarding the selection of CpG island and primer sequences.

Therefore, in addition to the new primers designed in this work, the same primer

sequences used in those articles were purchased. Since these primers mentioned before

were used in MSP techniques, they were considered as a possible alternative if the newly

designed primers could not be used. In fact, this was the case of CDKN2A(p16), for

which previously mentioned primer sequences were used instead. Moreover, both newly

design and Issa’s primer sequences for MINT31 failed to amplify correctly, and therefore,

new primer sequences were selected for this locus, based on the work by Weisenberger

et al (2006)125. The final primer sequences employed to test all samples by qMSP, and

their associated characteristics are summarized in Table 3.

Quantitative methylation-specific polymerase chain reaction (qMSP) Quantitative real-time methylation specific PCR was performed using

NZYSpeedy qPCR Green Master Mix with ROX (2X) (NZYTech, Lda., Lisbon, Portugal),

and β‑actin (ACTB) as the reference gene, to analyse CpG islands methylation levels of

CDKN2A(p16), MLH1 and MINT1, 2 and 31 promoters, in all tissue samples. Reactions

were carried out in 384-well plates using a LightCycler 480 instrument II (Roche,

Mannheim, Germany). Briefly, per each well 2 μL of modified DNA, 5 μL of Master Mix

and 0.3 μL of working primers’ solution 10 μM were added. Double-distilled water was

also added to complete the final volume (10 μL). To prepare working primers’ solutions,

10 μL of each front (F) and respective reverse (R) primer’s solution (100 μM) were diluted

in 180 μL of double-distilled water. The PCR program comprised a period of 3 minutes

at 95°C to activate the enzyme, followed by 45 cycles with 3 seconds at 95°C (for DNA

denaturation) and 30 seconds at a specific annealing temperature for each gene (for

annealing, extension and data acquisition) [see Table 3 for annealing temperature data].

An additional dissociation stage was included. All samples were run in triplicates and in

each plate five negative template controls were also run. Universal Methylated Human

DNA Standard (Zymo Research Corp., Irvine, CA, USA) was used to generate five serial

dilutions by a 5X dilution factor. These serial dilutions were run in each plate and were

used to generate a standard curve, thus allowing for absolute quantification and


49

determination of PCR efficiency. A run was considered valid when the slope of each

standard curve was above -3.60, corresponding to PCR efficiencies of > 90%, and R2

value of at least 5 relevant data points exceeded 0.96. The relative level of methylated

DNA for each gene/locus in each sample was determined using the formula: [(target

gene/(β-actin) x 1000]. Analysis of the qMSP data was performed using LightCycler®

480 software 1.5.0 SP3 (Roche, Mannheim, Germany).

To confirm the amplification of the specific product in standards and samples, the

melting curve and melting temperature data were analysed and only those samples

amplifying the specific product for each gene/locus were selected. For each of the

selected tumour samples, a specific gene/locus was proposed to be methylated if the

value of the previously described ratio was superior to any of the ratio values for the

selected control samples, considering the same gene/locus. When none of the control

samples amplified the specific product, all the selected tumour samples were proposed

to be methylated. However, only those samples with a ratio value greater than the

correspondent 25th percentile were considered to be methylated.

Statistical analysis Statistical analyses were performed using the statistical program SPSS software

(IBM SPSS® Statistics version 24.0, Chicago, IL, USA). All P-values were two-sided,

and statistical significance was set at P <0.05. Methylation of MLH1 was excluded from

all statistical analysis due to the small number of methylated cases. Categorical

clinicopathological and molecular variables were compared to CIMP status and

methylation of each marker using the Chi-square test or the Fisher’s exact test, as

applicable. Age was considered a categorical variable, and was further divided in two

groups according to median age, which was highly similar to mean age value. In addition,

the Chi-square analysis for N stage in MINT1 methylated tumours was replaced by

Fisher’s exact test because of the small number of methylated cases. For that purpose,

N3 and N4 stage groups were merged. Likewise, AJCC stage and specifically T stage

variables were clustered in two groups (lower and higher stages) to avoid small groups

to be analysed. Nonetheless, the analysis of some variables originated groups with less

than 5 elements, when analysing either the panel or the methylation of each marker.

Moreover, the clinicopathological factor tumour grade (G) was excluded from this initial

analysis, since almost all of the tumours were G2 (moderately differentiated), precluding

a correct statistical analysis for most of the studied groups.

Disease-specific survival time was measured from the date of diagnosis to the

date of death due to the progression of the disease, or the last clinical follow-up time for

surviving patients (censored). No patient from this cohort has died from other causes


apart from CRC. Disease-free survival time was measured from the date of surgery or

the last treatment performed (considering the patient was cured) to the date of

recurrence or the last clinical follow-up time (censored). In the case of multiple

recurrences, only the time elapsed by the first event was considered. Disease-specific

survival (DSS) and DFS were evaluated using log-rank statistic P-values for differences

in survival based on Kaplan-Meier’s approach (including graphical representations). Cox

proportional hazard regression model was used to calculate hazard ratios (HRs) of death

or recurrence according to clinical and molecular (KRAS, CIMP, methylation of markers)

features; and multivariable analysis was used to determine independent prognostic

factors. Due to the lack of enough representative cases, tumour grade was excluded

from the DFS analysis, and T3 and T4 tumours were joint in the same group. Likewise,

T1 and T2, as well as stage I and stage II or G1 and G2 tumours were combined in the

same group for all statistical tests. Moreover, throughout the descriptive text, P-values

mentioned correspond to Cox proportional hazard regression model.

Table 3 - List of primers’ sequences used and respective chromosomal location, size of the generated amplicon, temperature of annealing, GenBank Accession number and specific location in the accessed sequence.

Gene or locus

Chrom. location Sequence (5’–3’) Size,

bp T

Annealing, ºC

GenBank Accession

No. Location, bp

ACTB-F ACTB-R 7p22.1 TGGTGATGGAGGAGGTTTAGTAAGT

AACCAATAAAACCTACTCCTCCCTTAA 133 60 Y00474 390–522

CDKN2A-F CDKN2A-R 9p21 TTATTAGAGGGTGGGGCGGATCGC

GACCCCGAACCGCGACCGTAA 150 65 AF527803 19906–20056

MINT1-F MINT1-R 5q13-14 GGAGAGTAGGGGAGTTCGC

CTTCGCCTAACCTAACGC 119 62 AF135501 212-331

MINT2-F MINT2-R 2p22-21 TTTAGTATTTAAGTTCGTTGGC

ACGATTCCGTACGCCTTT 117 60 AF135502 431–548

MINT31-F MINT31-R 17q22 GTCGTCGGCGTTATTTTAGAAAGTT

CACCGACGCCCAACACA 72 60 AC021491 50059-50131

MLH1-F MLH1-R 3p21.3 GTAGTCGTTTTAGGGAGGGAC

TCAATACCTCGTACTCACGTTC 156 64 AY217549 1750-1906


51

RESULTS

PROJECT I

Considering tumour samples, the group of selected patients included nine men

and one woman, with a median age of 64 years old (range 53-67). Two patients were

diagnosed with AJCC stage I, two as stage II, three as stage III, and three as stage IV.

All patients had adenocarcinomas; in eight patients the tumour was localized in the colon,

while two patients had rectal cancer. In nine patients the tumour was of moderately

differentiated grade (G2), but poorly differentiated (G3) in the other patient. One patient

with colon cancer received neoadjuvant therapy (RT) before surgery.

When no correction to multiple t-tests was applied, a few significant changes of

expression were spotted [Tables 4 and 5]. However, the profiling analysis of 90 lncRNAs

revealed that no transcript was differentially expressed between any pair of the four

groups compared, after Holm-Šídák correction. Likewise, no difference was found

comparing all tumour samples with healthy mucosa equivalents, or pitting all samples

with lower BER repair capacity against samples presenting higher BER repair capacity

[Table 6]. Nevertheless, those previous results, depicted in Table 4 and 5, will be

described herein, albeit bearing in mind the loss of significance after employing the

correction model.

After comparing each pair of the four groups, fifteen different transcripts were

found to be up or down-regulated. SNHG4, LUST, GAS5-family of transcripts, E2F4

antisense, anti-NOS2A and BACE1AS family of transcripts were all found to be down-

regulated in lower BER repair capacity tumours compared to healthy mucosa samples

with the opposite behaviour. GAS5 (family) and E2F4 antisense transcripts were also

found to be down-regulated in TH group compared to HH group, while MEG9 was found

to be up-regulated. Except from mascRNA, all down-regulated transcripts in TL group

were commonly affected in TH group, when each of these two groups was compared to

HL group. Likewise, IGF2AS family of transcripts expression was found to be altered in

both groups of tumours compared with HL group, but in this case the transcript was up-

regulated. Furthermore, the biggest fold change was reported for IGF2AS transcripts.

The analysis of repair capacity within tumour or healthy mucosa groups of samples

revealed seven transcripts down-regulated in healthy mucosa with the increase of the

repair capacity, but no differences were found comparing TH and TL groups [Table 4].


Table 4 - Long noncoding RNAs differentially expressed between the four groups of samples formed HH, HL, TH and TL, before Holm-Šídák correction. Significant P-values (P<0.05) not adjusted are represented below the Fold

Change values. No lncRNA was found to be differently expressed between TL and TH groups. After adjustment of P-

value according to Holm-Šídák correction for multiple comparisons, no significant differences were found. HH: Healthy

mucosa with Higher levels of BER repair capacity, HL: Healthy mucosa with Lower levels of BER repair capacity, TH:

Tumour with Higher levels of BER repair capacity, TL: Tumour with Lower levels of BER repair capacity.

LncRNAs HH HL HL

TH TL TL TH HH

H19 antisense -3.10 (0.043)

-3.50 (0.023)

Zfas1 -7.01 (0.020)

-12.27 (0.013)

-7.61 (0.017)

SNHG4 -2.58 (0.018)

SAF -4.83 (0.045)

-11.38 (0.020)

-7.07 (0.028)

HOTAIRM1 -7.07 (0.024)

-9.34 (0.020)

-8.43 (0.021)

IGF2AS (family) 41.22 (0.003)

17.90 (0.025)

RNCR3 -14.32 (0.033)

-24.79 (0.029)

-12.98 (0.035)

LUST -7.17 (0.003)

GAS5-family -4.03 (0.015)

-2.68 (0.005)

E2F4 antisense -5.49 (0.010)

-1.52 (0.005)

anti-NOS2A -2.07 (0.024)

BACE1AS (family) -1.83 (0.049)

Jpx -14.42 (0.021)

-7.04 (0.032)

-11.40 (0.023)

mascRNA -5.69 (0.043) -4.81

(0.048)

MEG9 2.58 (0.017) -12.05

(0.047)

Additionally, Zfas1, H19 antisense and SNHG4 persisted as down-regulated

transcripts after comparing all tumours with all healthy mucosa samples, irrespective of

repair capacity. In the same manner, Zfas1, SAF and HOTAIRM1 were not only down-

regulated comparing HH to HL groups, but also when all samples with higher BER repair

capacity (both tumours and healthy mucosa) were compared with all samples with lower

repair capacity. Moreover, two additional transcripts (ST7OT and lincRNA-p21) were

also found to be down-regulated in tumour samples [Table 5].


53

Table 5 - Long noncoding RNAs differentially expressed between Healthy mucosa and Tumour samples, and samples with Lower and High BER repair capacity, before Holm-Šídák correction. Significant P-values (P<0.05) not

adjusted are represented after the Fold Change values. After adjustment of P-value according to Holm-Šídák correction

for multiple comparisons, no significant differences were found.

LncRNAs Healthy vs. Tumour Lower vs. Higher

H19 antisense -2.72 (0.015)

Zfas1 -5.05 (0.038) -5.37 (0.033)

SNHG4 -5.63 (0.044)

SAF -5.26 (0.031)

HOTAIRM1 -5.06 (0.038)

ST7OT -7.89 (0.018)

lincRNA-p21 -2.67 (0.026)

Table 6 - P-values for the differential expression of long noncoding RNAs between the four groups of samples formed HH, HL, TH and TL, and between Healthy mucosa and Tumour samples or samples with Lower and High BER repair capacity, after Holm-Šídák correction. After adjustment of P-value for the results described previously, no

significant differences were found.

LncRNAs HH HL HL Healthy vs.

Tumour

Lower vs.

Higher TH TL TL TH HH

H19 antisense 0.976 0.868 0.749

Zfas1 0.835 0.701 0.793 0.966 0.949

SNHG4 0.788 0.979

SAF 0.978 0.833 0.916 0.940

HOTAIRM1 0.876 0.833 0.849 0.968

IGF2AS (family) 0.269 0.886

RNCR3 0.943 0.916 0.952

LUST 0.250

GAS5-family 0.748 0.348

E2F4 antisense 0.599 0.381

anti-NOS2A 0.875

BACE1AS (family) 0.987

Jpx 0.849 0.935 0.877

mascRNA 0.976 0.984

MEG9 0.789 0.984

ST7OT 0.801

lincRNA-p21 0.905


PROJECT II

Patients’ characteristics and CpG island methylation at specific loci Of the 211 CRC cases, 34.1% (n=72) were females and 65.9% (n=139) were

males, with a median age at diagnosis of 61 years (61.5 years for women and 60.0 years

for men). Almost half of all cases (49.3%) were identified in the rectum (n=104), while

34.1% were found in the distal colon (n=72) and 16.6% in the proximal colon (n=35).

Moreover, 39.3% of all tumours were KRAS mutated (n=83). Regarding stage, one CRC

case was T1, 21 cases were T2, 170 were T3 stage and 16 were T4 staged. Therefore,

88.2% of all cases were in a locally advanced stage (T3&T4). Most of the cases displayed

lymph node metastasis, 30.4% at a N1 stage (n=64) and 31.3% at a N2 stage (n=66);

and 50.7% (n=107) displayed distant metastasis at the time of diagnosis. Accordingly,

most patients were also diagnosed with advanced AJCC stages of the disease. In fact,

more than half of all patients (n=107) were diagnosed with stage IV, 24.6% with stage III

and the remaining 23.7% with stages I or II. In addition, a fraction of all patients (32.7%)

was submitted to neoadjuvant therapy, whereas the majority of the studied patients

received adjuvant therapy at some point during the follow-up (80.1%). The main

clinicopathological and molecular variables of the 211 selected CRC cases are depicted

in Table 7.

CIMP status was evaluated by the quantitative method SYBR® Green-based

qMSP of a five genes/loci panel previously reported. MINT31 showed the highest

methylation frequency, whereas MLH1 displayed the lowest, with 15.2% and 0.9%,

respectively. Methylation frequencies of the remaining genes were 6.6% for MINT1,

14.7% for MINT2 and 11.4% for CDKN2A(p16). Significant positive associations were

found among MINT1, MINT2, MINT31, and CDKN2A(p16) methylation levels, suggestive

of a hypermethylator phenotype (CIMP) in a subset of cases [Fig.5]. Since MLH1 was

found to be methylated only in two cases, no statistical analysis was performed. When

methylation of all genes/loci was grouped based on methylation of 0 or 1 marker versus

>1 marker for the CIMP phenotype, 18 patients were defined as CIMP positive (8.5%)

and 193 patients were defined as CIMP negative (91.5%). In a trichotomous

categorization model, 136 patients were classified as CIMP-0, 72 patients as CIMP-L

and three patients as CIMP-H (two patients with 4 methylated markers and one patient

with all 5 markers methylated). Regarding CIMP-L patients, 57 only presented one

methylated gene, while 12 were methylated in 2, and 3 patients displayed methylation in

3 genes [Fig.5].

Two of all patients had no information regarding tumour invasion depth (T), lymph

node metastasis (N) and distant metastasis (M). Therefore, those patients had no AJCC


55

stage information. Additionally, four other patients had no N stage information, and for

14 of all patients KRAS mutation status was not determined. In 211 patients with

clinicopathological and molecular characteristics available for analyses, no statistically

significant differences were found between CIMP+ and CIMP– tumours. However, CIMP-

L cases were more likely to present ≥4 regional lymph nodes metastasis (P=0.018), and

strongly associated with mutated KRAS (P<0.001) [Fig.6][Table 7]. Similarly to CIMP+/–

status, individual methylation status of MINT2 and CDKN2A(p16) did not associate with

any of the studied variables. However, MINT31 methylation associated with mutated

KRAS (P=0.004) [Fig.6], while MINT1 methylation associated with the absence of

regional lymph nodes metastasis (P=0.017) [Table 8] [Appendix I].

Due to the small number of CIMP-H cases, this category was excluded from the

statistical analysis. However, worth of mention, two of the CIMP-H tumours were

biopsied from female patients and were IIA (T3N0M0) staged, while the male patient was

diagnosed with a IVA (T3N0M1a) stage tumour. All three patients were older than 61.

Moreover, none of the three tumours were located in the rectum – two tumours were

found in the distal colon and one in the proximal colon –, and only one tumour presented

mutated KRAS (data not shown).

The same statistical analysis performed excluding all patients receiving

neoadjuvant therapy revealed no major differences. Indeed, CIMP-L continued to be

significantly associated with mutated KRAS (P=0.009) and N2 stage (P=0.002).

Likewise, MINT31 methylated cases presented a strong tendency towards KRAS

mutated (P=0.054). However, neither MINT1 nor MINT2 methylation displayed a

significant association with N1&2 stage after Fisher’s exact test was performed,

considering only those 141 patients who did not receive any neoadjuvant treatment

[Appendix II and III]. Of mention, no information concerning neoadjuvant therapy was

available for one of the 211 patients.

The analysis of KRAS mutations revealed that almost all of these were located in

the second exon of the gene – one was located in exon 3 and another sample presented

a mutation in exon 4 of the KRAS gene. Furthermore, one mutated KRAS sample had

no further information regarding the specific exon altered. Considering only mutations in

the second exon of the gene, nine different single mutations affecting mostly Glycine 12

or Glycine 13 were found. However, none of the described substitutions of Glycine by

another residue were correlated with any of the two CIMP categorizations or methylation

of any marker, particularly CIMP-L and methylation of MINT31 (data not shown).


Fig.5 - Performance of the classic CIMP panel. The 211 tumours were screened against the classic set of CIMP

markers. The alignment of each tumour is maintained across all analysis. At the left side: dichotomous heat maps

representing DNA methylation data for all 5 genes/loci (red: methylayed, light blue: unmethylated), and resultant CIMP

categorization according to both dichotomous (black: CIMP+, light grey: CIMP–) and trichotomous (red: CIMP-H, black:

CIMP-L, light grey: CIMP-0) models. At the right side: histogram of the methylation frequency distribution for the set of

classic CIMP markers.


57

Table 7 - Distribution of clinicopathological and molecular variables for all CRC patients and association with CIMP status. Number of cases (and respective percentage) distributed per each category from all CRC cases (N=211),

excluding tumour grade. Dichotomous (positive/negative) and trichotomous (CIMP-0, CIMP-Low and excluding CIMP-

High) CIMP categorization, and its distribution and association with all represented variables. P-values were calculated

using the Chi-squared test or the Fisher’s exact test. Significant P-values (P<0.05) are represented in bold.

*P-Value calculated with >20% cells having expected counts less than 5.

Variables Cases (%) CIMP- CIMP+ CIMP-0 CIMP-L No. % No. % No. % No. % 193 91.5 18 8.5 136 64.5 72 34.1 Gender

Female Male

72 (34.1) 139 (65.9)

64

129

33.2 66.8

8

10

44.4 55.6

44 92

32.4 67.6

26 46

36.1 63.9

P=0.437 P=0.644 Age at diagnosis

≤61 (median age) >61

114 (54.0) 97 (46.0)

108 85

56.0 44.0

6

12

33.3 66.7

75 61

55.1 44.9

39 33

54.2 45.8

P=0.084 P=1.000 Tumour location

Rectum Distal Colon Proximal Colon

104 (49.3) 72 (34.1) 35 (16.6)

92 68 33

47.7 35.2 17.1

12 4 2

66.7 22.2 11.1

66 51 19

48.5 37.5 14.0

38 29 15

52.8 26.4 20.8

P=0.304* P=0.197 KRAS status

Wild-type Mutated ND

114 (54.1) 83 (39.3) 14 (6.6)

108 72 13

64.7 35.3

-

6

11 1

40.0 60.0

-

87 42 7

67.4 32.6

-

25 40 7

38.5 61.5

- P=0.070 P<0.001 AJCC stage

I&II III&IV ND

50 (23.7) 159 (75.4) 2 (0.9)

45

146 2

23.6 76.4

-

5

13 -

27.8 72.2

-

30

105 1

22.2 77.8

-

18 53 1

25.4 74.6

- P=0.773 P=1.000 Tumour invasion depth (T)

T1&T2 T3&T4 ND

22 (10.4) 186 (88.2) 3 (1.4)

19

172 3

9.90 90.1

-

3

15 -

16.7 83.3

-

15

120 2

11.1 88.9

-

7

64 1

9.90 90.1

- P=0.413 P=1.000 Lymph node metastasis (N)

N0 N1 N2 ND

75 (35.5) 64 (30.4) 66 (31.3) 6 (2.8)

67 62 58 6

35.8 33.2 31.0

-

8 2 8 -

44.4 11.2 44.4

-

48 49 35 4

36.4 37.1 26.5

-

24 15 31 2

34.3 21.4 44.3

- P=0.149 P=0.018 Distant Metastasis (M)

M0 M1 ND

102 (48.4) 107 (50.7) 2 (0.9)

94 97 2

49.2 50.8

-

8

10 -

44.4 55.6

-

65 70 1

48.1 51.9

-

35 36 1

49.3 50.7

- P=0.807 P=0.885 Neoadjuvant therapy

Yes No ND

69 (32.7) 141 (66.8) 1 (0.5)

62

130 1

32.3 67.7

-

7

11 -

38.9 61.1

-

43 92 1

31.9 68.1

-

26 46 -

36.1 63.9

- P=0.604 P=0.540 Adjuvant therapy

Yes No

169 (80.1) 42 (19.9)

154 39

79.8 20.2

15 3

83.3 16.7

108 28

70.4 20.6

59 13

81.9 18.1

P=1.000 P=0.717


Table 8 - Association between clinicopathological and molecular variables and each of the five genes/loci constituting the classic CIMP panel. Classification of tumours according to the methylation status of each gene or locus

(MINT1, MINT2, MINT31 and CDKN2A(p16)), and its distribution and association with all represented variables. P-values

were calculated using the Chi-squared test or the Fisher’s exact test. Significant P-values (P<0.05) are represented in

bold. To avoid oversizing of the table, non-significant results for tumour invasion depth (T), distant metastasis (M),

neoadjuvant and adjuvant therapies variables were excluded from the represented table. A complete version of this table

is depicted in Appendix I.

Variables MINT1 MINT2 MINT31 CDKN2A(p16)

M UM M UM M UM M UM

No. % No. % No. % No. % No. % No. % No. % No. %

14 6.6 197 93.4 31 14.7 180 89.6 32 15.2 179 84.8 24 11.4 187 88.6

Gender Female Male

8 6

57.1 42.9

64

133

32.5 67.5

11 20

35.5 64.5

61

119

33.9 66.1

14 18

43.8 56.3

58

121

32.4 67.6

7

17

29.2 70.8

65

122

34.8 65.2

P=0.080 P=0.841 P=0.229 P=0.654

Age ≤61 >61

6 8

57.1 42.9

108 89

54.8 45.2

12 19

38.7 61.3

102 78

56.7 43.3

16 16

50.0 50.0

98 81

54.7 45.3

11 13

45.8 54.2

103 84

55.1 44.9

P=0.418 P=0.079 P=0.701 P=0.515

Location Rectum Distal Proximal

6 5 3

42.9 35.7 21.4

98 67 32

49.7 34.0 16.3

18 8 5

58.1 25.8 16.1

86 64 30

47.8 35.6 16.7

14 9 9

43.8 28.1 28.1

90 63 26

50.3 35.2 14.5

14 6 4

58.3 25.0 16.7

90 66 31

48.1 35.3 16.6

P=0.840* P=0.520 P=0.160 P=0.574

KRAS WT Mutated ND

5 8 1

38.5 61.5

-

109 75 13

59.2 40.8

-

12 17 2

41.4 58.6

-

102 66 12

60.7 39.3

-

10 20 2

33.3 66.7

-

104 63 12

62.3 37.7

-

9

12 3

42.9 57.1

-

105 71 11

59.7 40.3

-

P=0.158 P=0.067 P=0.004 P=0.164

AJCC stage I&II III&IV ND

5 8 1

38.5 61.5

-

45

151 1

23.0 77.0

-

9

22 -

29.0 71.0

-

41

137 2

22.6 77.4

-

10 22 -

31.3 68.7

-

40

137 2

22.6 77.4

-

5

19 -

24.9 75.1

-

45

140 2

24.3 75.7

-

P=0.310 P=0.496 P=0.367 P=0.804

N N0 N1/(N1&2) N2 ND

9 4 - 1

69.2 30.8

- -

66

126 - 5

34.4 65.6

- -

11 5

14 1

36.6 16.7 46.7

-

64 59 52 5

36.6 33.7 29.7

-

14 6

12 -

43.3 18.8 37.5

-

61 58 54 6

35.3 33.5 31.2

-

8 5

11 -

25.0 29.2 45.8

-

67 59 55 6

37.0 32.6 30.4

-

P=0.017 P=0.097 P=0.253 P=0.275

*P-Value calculated with >20% cells having expected counts less than 5. M, methylated; UM, unmethylated; WT, wild-type.


59

Fig.6 - Comparison between the classic CIMP panel, MINT31 methylation and KRAS mutation status. The 211

tumours were screened against the classic set of CIMP markers. The alignment of each tumour is maintained across all

analysis. At the left side: simplified heat maps representing trichotomous CIMP categorization (red: CIMP-H, black: CIMP-

L, light grey: CIMP-0), MINT31 methylation (red: methylayed, light blue: unmethylated), and KRAS mutation status (green:

wild-type, blue: mutated, white: not determined). Right side: relative frequencies of KRAS mutation for CIMP-0 and CIMP-

L tumours, with colour codings as described above. WT, wild-type; M, mutated; ND, not determined.

Prognostic factors for survival: disease-specific survival The DSS for one of the 211 CRC patients could not be determined. The median

follow-up of all 210 CRC patients was 52 months (range: 5–212 months). At the time of

the last follow-up, 14 patients were alive with no evidence of cancer, 11 patients were

alive with cancer progression, while the remaining 185 patients had deceased (due to

CRC progression). The DSS rate of the 210 patients was 99.5%, 40.7% and 5.5% at

one, five and ten years of follow-up, respectively. Univariable survival analysis showed

a significant association between a decrease in DSS and an older age at diagnosis

(P=0.005) [Fig.7A], locally advanced tumour stage (T4) (P<0.001), ≥4 regional lymph


node metastasis (N2) (P=0.024) and distant metastasis (P<0.001), as well as with AJCC

tumour stage IV (P<0.001) [Fig.7B] and neoadjuvant therapy (P=0.040) [Fig.7C]. The

remaining clinicopathological and molecular parameters presented no prognostic value

as they were not significantly associated with DSS. However, a trend towards decreased

DSS was reported for less differentiated tumours (G3) (P=0.066). Importantly, the total

number of G3 tumours was critically small and thus, this result shall be interpreted

cautiously [Table 9].

In a multivariable analysis, using Cox proportional hazard regression model, older

age at diagnosis, higher AJCC tumour stage (IV) and neoadjuvant therapy were

independently associated with decreased DSS (P=0.034, P<0.001 and P=0.040,

respectively) [Table 9]. Indeed, the risk of death due to the progression of the disease

for older patients and patients submitted to neoadjuvant therapy was, respectively, 1.387

(95% CI 1.024-1.877) and 1.406 (95% CI 1.032-1.914) times higher, while patients

diagnosed with stage IV tumours had an increased risk of 1.887 (95% CI 1.309-2.719,

P=0.001) and 1.912 (95% CI 1-332-2.745, P<0.001) times relative to stages I&II and

stage III tumours, respectively. T, N and M individual stages were intentionally excluded

from the multivariable analysis as each of these are inherent to the AJCC staging, by

definition, and may be considered linearly dependent by the software of analysis

(particularly M stage).

According to the general analysis, including all 210 CRC cases, neither CIMP+/-

(1.192 95% CI 0.732-1.941, P=0.481) nor CIMP-0/L (0.952 95% CI 0.700-1.294,

P=0.753) categorization of tumours displayed significant differences in DSS [Table 9]

[Fig.7E,F]. Likewise, after exclusion of all 68 patients treated with neoadjuvant therapy,

no significant association with CIMP panel was found [Appendix IV].

Considering only CIMP+ cases, metastatic tumours were significantly associated

with worse outcome (P=0.003), and tumours at stages III&IV presented a slight trend

towards worse prognosis (P=0.080) [Appendix V]. Once more, this result shall be

interpreted cautiously due to the small number of CIMP+ cases. In CIMP– group, the

variables independently associated with worse prognosis were the same as in the

general analysis, as this group contains almost all cases (not shown). In CIMP-0

tumours, proximal colon localization (P=0.001) and AJCC tumour stage IV (P=0.008)

were independently associated with a decrease in DSS; while in CIMP-L tumours,

neoadjuvant therapy (P=0.018) and mutated KRAS (P=0.002) were independently

associated with worse and better outcome, respectively [Appendix V]. Although no

statistical analysis was performed for CIMP-H, all three patients eventually deceased

from disease progression, with DSS times of circa 86, 55 and 27 months, respectively

(data not shown).


61

Methylation of MINT loci and CDKN2A(p16) were closely interrelated, and it was

considered possible that the adverse effects of methylation of one member of the group

could be incorrectly ascribed to tumours displaying a generalized methylator phenotype.

Therefore, the prognostic significance of each of the four markers was independently

examined. Methylation of each MINT loci did not associate with disease outcome. Only

CDKN2A(p16) methylation associated with worse outcome in a univariable analysis

(1.578 95% CI 1.016-2.450, P=0.042) [Table 9] [Fig.8D]. However, the significance was

lost in a multivariable analysis, although borderline (HR 1.561 95% CI 0.999-2.440,

P=0.051) [Table 9]. After exclusion of all 68 patients subjected to neoadjuvant therapy,

CDKN2A(p16) methylation was independently associated with worse prognosis (HR

1.838, 95% CI, 1.090-3.097, P=0.022) [Appendix IV].

Table 9 - Univariable and multivariable prognostic analyses: disease-specific survival analysis of CRC patients according to represented variables and CIMP panel/markers methylation. One patient had no information regarding

the estimation of DSS. Multivariable analysis was performed considering only those variables presenting a P-value<0.05

in the univariable analysis (excluding T, N and M stages). Significant P-values (P<0.05) are represented in bold.

Variables Univariable analysis Multivariable analysis

Median (mo) (95 % CI) Pa HR (95 % CI) Pb HR (95 % CI) Pb

Gender Female (72) Male (138)

55.5 (43.8-56.8) 50.3 (48.1-62.9)

0.998 1 (referent)

1.000 (0.735-1.359)

-

0.998

Age ≤61 (113) >61 (97)

57.1 (49.6-64.7) 48.8 (42.1-55.5)

0.005 1 (referent)

1.520 (1.132-2.042)

-

0.005

1 (referent)

1.387 (1.024-1.877)

-

0.034 Location

Rectum (103) Distal (72) Proximal (35)

55.5 (45.3-65.7) 55.8 (48.8-62.8) 44.7 (29.6-59.7)

0.223 0.715 (0.476-1.072) 0.710 (0.460-1.096)

1 (referent)

0.226 0.104 0.122

-

KRAS WT (113) Mutated (83)

51.6 (45.7-57.5) 54.2 (46.4-62.0)

0.684 1 (referent)

0.938 (0.690-1.275)

-

0.684

AJCC stage I & II (50) III (51) IV (107)

59.6 (56.8-62.4) 66.4 (57.3-75.5) 41.6 (36.9-46.3)

<0.001 0.530 (0.368-0.764) 0.523 (0.364-0.751)

1 (referent)

<0.001 0.001

<0.001 -

0.534 (0.379-0.771) 0.523 (0.364-0.753)

1 (referent)

<0.001 0.001

<0.001 -

T T1&T2 (22) T3 (170) T4 (16)

63.7 (57.1-70.4) 53.7 (48.7-58.7) 30.7 (29.2-32.2)

<0.001 0.300 (0.155-0.583) 0.307 (0.186-0.539)

1 (referent)

<0.001 <0.001 <0.001

-

N N0 (75) N1 (63) N2 (66)

58.4 (52.9-64.0) 59.1 (42.9-75.3) 43.6 (33.3-53.9)

0.023 0.649 (0.458-0.922) 0.649 (0.447-0.941)

1 (referent)

0.024 0.016 0.023

-

M M0 (101) M1 (107)

63.3 (59.4-67.2) 41.6 (36.9-46.3)

<0.001 1 (referent)

1.899 (1.413-2.554)

-

<0.001


Grade

G1&2 (125) G3 (5)

57.1 (50.6-63.7) 43.4 (19.8-66.9)

0.058 1 (referent)

2.342 (0.944-5.807)

-

0.066

Neoadjuvant Yes (68) No (141)

47.6 (40.3-54.9) 55.8 (50.0-61.7)

0.040 1.377 (1.014-1.870)

1 (referent)

0.040

-

1.406 (1.032-1.914)

1 (referent)

0.040

- Adjuvant

Yes (168) No (42)

50.1 (44.5-55.8) 59.3 (53.0-65.7)

0.083 1.381 (0.957-1.993)

1 (referent)

0.085

-

CIMP Positive (18) Negative(192)

51.9 (47.0-56.8) 55.3 (36.3-54.2)

0.481 1.192 (0.732-1.941)

1 (referent)

0.481

-

CIMP CIMP-0 (136) CIMP-L (71)

51.0 (44.1-57.9) 54.8 (46.2-63.5)

0.753 1 (referent)

0.952 (0.700-1.294)

-

0.753

MINT1 M (14) UM (196)

50.7 (36.7-64.8) 52.3 (47.2-57.4)

0.926 1.027 (0.592-1.779)

1 (referent)

0.926

-

MINT2 M (31) UM (179)

60.4 (53.1-67.6) 51.6 (46.6-56.7)

0.969 0.992 (0.662-1.486)

1 (referent)

0.969

-

MINT31 M (32) UM (178)

60.4 (45.0-75.7) 50.7 (44.7-56.6)

0.199 0.768 (0.513-1.150)

1 (referent)

0.200

-

P16 M (23) UM (187)

44.7 (31.4-57.9) 53.4 (47.7-59.7)

0.040 1.578 (1.016-2.450)

1 (referent)

0.042

-

1.561 (0.999-2.440)

1 (referent)

0.051

- a Log-rank test b Cox proportional hazard regression model CI, confidence interval; HR, hazard ratio; M, methylated; UM, unmethylated; WT, wild-type.

Deepening the prognostic analysis, some significant associations were found

after stratifying the test for each CIMP marker. In MINT31 methylated cases, mutated

KRAS was independently associated with better prognosis (P=0.015). Similarly, in

CDKN2A(p16) or MINT2 methylated tumours, an independent association between

decreased survival and AJCC stage IV tumours was also found (relative to stage I&II

(P=0.041) or III tumours (P=0.024), respectively) [Appendix VI].

Stratifying the analysis by the other molecular and clinicopathological parameters

revealed additional significant associations between CIMP panel or individual markers

with DSS. For instance, CDKN2A(p16) methylation associated with worse outcome in

male patients (P=0.042), or patients with 61 years old or younger (P=0.024). In proximal

tumours CIMP-L was associated with better prognosis (P=0.013) [Appendix VII],

whereas in KRAS wild-type tumours CIMP-L was instead associated with worse

prognosis (P=0.010). Moreover, in tumours presenting KRAS mutation, both CIMP-L and

MINT31 methylation were associated with better outcome (P=0.015 and P=0.029,

respectively). In the subgroup of patients without adjuvant treatment, MINT31

methylation was associated with better prognosis (P=0.046) [Appendix VIII].


63

Fig.7 - Kaplan-Meier curves analysis for disease-specific survival according to age at diagnosis, AJCC tumour stage, neoadjuvant therapy, CIMP panel and CDKN2A(p16) methylation status. All four variables that were found to

be associated with worse DSS after the multivariable analysis were screened through the Kaplan-Meier survival plot. A:

older age at diagnosis was associated with reduced DSS; B: AJCC stage IV tumours were associated with reduced DSS;

C: neoadjuvant therapy was associated with reduced DSS; D: methylated status of CDKN2A(p16) marker was associated

with reduced DSS. No differences in DSS time between CIMP negative and positive tumours as well as between CIMP-

0 and CIMP-Low tumours were found (E and F, respectively). Represented P-values were calculated by the Log-rank

test.

A

C

B

D

E F


Prognostic factors for survival: disease-free survival

Of the 211 CRC patients, only 109 patients (51.6%) were used for the DFS

analysis. The median DFS time of the 109 CRC patients was 16 months (range: 2–111

months). At the time of the last follow-up, 4 patients were alive with no evidence of

recurrence, while the remaining 105 patients had at least one recurrence (9 of these

patients had a second recurrence or more). The DFS rate for the 109 patients was

70.6%, 18.6% and 6.6% after one year, three and five years of follow-up, respectively.

Survival analysis showed a significant independent association between male gender

and shorter DFS (HR 1.549, 95% CI 1.030-2.329, P=0.035) [Fig.8A]. None of the

remaining variables significantly associated with DFS [Table 10].

Neither the CIMP+/– panel (0.554 95% CI 0.241-1.275, P=0.161) nor individual

CIMP markers significantly associated with DFS [Table 10] [Fig.8B,C,D]. Moreover, a

small number of tumours presented CIMP+ or MINT1 methylation and thus,

interpretation of the results shall be attended cautiously. Moreover, considering only

CIMP– or CIMP-L patients, none of the parameters associated with DFS. However, in

CIMP-0, proximal tumours were independently associated with shorter DFS (P=0.008)

(data not shown). Considering CIMP-H cases, two of the three patients were never cured

during the follow up time, whereas the other patient had a disease-free survival time of

64 months (data not shown).

Table 10 - Univariable prognostic analyses: disease-free survival analysis for CRC patients according to represented variables and CIMP panel/markers methylation. A total of 109 patients (51.6%) were considered for the

analysis of DFS. Significant P-values (P<0.05) are represented in bold.

Variables Univariable analysis

Median (mo) (95 % CI) Pa HR (95 % CI) Pb

Gender Female (39) Male (70)

20.6 (12.6-28.6) 14.7 (12.3-17.0)

0.034 1 (referent)

1.549 (1.030-2.329)

-

0.035 Age

≤61 (63) >61 (46)

16.4 (13.2-19.5) 16.8 (10.5-23.1)

0.456 1 (referent)

1.160 (0.785-1.714)

-

0.457 Location


16.1 (10.1-22.1) 21.1 (12.0-30.2) 9.99 (7.40-12.6)

0.126 0.839 (0.493-1.427) 0.575 (0.520-1.033)

1 (referent)

0.130 0.517 0.064

- KRAS

WT (59) Mutated (42)

16.8 (13.6-20.1) 14.3 (11.0-17.5)

0.505 1 (referent)

1.149 (0.763-1.731)

-

0.506 AJCC stage

I & II (46) III (39) IV (23)

17.6 (9.97-25.3) 16.8 (13.7-20.0) 13.2 (9.05-17.3)

0.463 1.033 (0.603-1.770) 1.320 (0.756-2.303)

1 (referent)

0.465 0.906 0.329

-


65

T

T1&T2 (15) T3&T4 (93)

24.4 (18.7-30.0) 15.9 (13.1-18.8)

0.249 0.718 (0.407-1.265)

1 (referent)

0.251

- N

N0 (51) N1 (32) N2 (23)

16.4 (9.93-22.8) 17.2 (4.86-29.6) 15.3 (12.0-18.5)

0.183 0.981 (0.619-1.555) 1.546 (0.933-1.560)

1 (referent)

0.188 0.936 0.091

- M

M0 (85) M1 (23)

16.9 (12.1-21.7) 13.2 (9.05-17.3)

0.597 1 (referent)

0.873 (0.529-1.443)

-

0.597 Neoadjuvant

Yes (31) No (78)

14.5 (11.2-17.9) 17.2 (13.1-21.3)

0.481 1.165 (0.761-1.784)

1 (referent)

0.481

- Adjuvant

Yes (68) No (41)

16.8 (14.0-25.0) 16.4 (11.7-21.8)

0.812 0.953 (0.643-1.414)

1 (referent)

0.812

- CIMP

Positive (6) Negative(103)

30.6 (4.18-57.0) 16.1 (13.8-18.4)

0.159 0.554 (0.241-1.275)

1 (referent)

0.161

- CIMP

CIMP-0 (70) CIMP-L (38)

16.8 (12.1-21.6) 14.3 (9.40-19.1)

0.848 1 (referent)

1.041 (0.691-1.566)

-

0.848 MINT1

M (6) UM (103)

16.1 (0.00-59.9) 16.8 (14.1-19.5)

0.076 0.442 (0.175-1.115)

1 (referent)

0.084

- MINT2

M (13) UM (96)

20.5 (3.92-37.0) 16.4 (14.1-18.6)

0.478 0.808 (0.447-1.459)

1 (referent)

0.479

- MINT31

M (15) UM (94)

17.2 (7.06-27.4) 16.4 (13.9-18.8)

0.546 1.843 (0.484-1.468)

1 (referent)

0.546

- P16

M (13) UM (96)

19.9 (10.8-29.0) 16.4 (14.1-18.6)

0.887 1.043 (0.582-1.870)

1 (referent)

0.887

- a Log-rank test b Cox proportional hazard regression model CI, confidence interval; HR, hazard ratio; M, methylated; UM, unmethylated; WT, wild-type.


Fig.8 - Kaplan-Meier curves analysis for disease-free survival according to gender, CIMP panel and CDKN2A(p16) methylation status. Male gender was found to be associated with decreased DFS time (A), but no

association was found analysing the methylation status of CDKN2A(p16) marker (B). No differences in DFS time between

CIMP negative and positive tumours as well as between CIMP-0 and CIMP-Low tumours were found (C and D,

respectively). Represented P-values were calculated by the Log-rank test.

A B

C D


67

DISCUSSION CRC is one of the most common malignancies in the world, and although

screening for early detection of CRC has the potential to reduce both the incidence and

mortality of the disease, still the overall survival rate has not changed dramatically, and

a large number of individuals will develop CRC each year and eventually decease

following disease progression.3 Development of CRC is a complex biological process,

involving multiple genomic and epigenomic alterations.137 In fact, intensive investigation

over the last few decades have focused on the comprehension of genomic mechanisms,

and particularly the role of protein-coding genes in the pathogenesis of CRC.54 It is now

time to explore new horizons that may very well represent the target for future treatment

options or diagnostic tools; and when the subject is not directly the genome, it is

inevitably epigenetics, in its wide and complex web of regulatory processes involving our

genetic material. The last few years have been essential to definitely prove the

importance of epigenetics in cancer development and treatment. Therefore, it was the

selected matter in this work.

PROJECT I

Recent studies of lncRNAs have highlighted the importance of this new class of

the non-coding part of our genome. Enormous amount, diversity of functions and great

flexibility may be the explanation for their commonly deregulated expression, which is

often significantly correlated with carcinogenesis.64

CRC has long been associated with defects in DNA repair, mostly with genetic

alterations and aberrant DNA methylation of MMR genes. In contrast, BER and NER

pathways are not described as significantly related with CRC development and outcome

in most genetic studies published. However, more intense research is needed to achieve

a better understanding of these repair mechanisms in the particular case of CRC, and

completely rule out an important role played by either BER or NER pathways in the

development of the disease. In an effort to help tackling the problem, the first project

presented here was dedicated to the epigenetic study of the BER repair pathway in

sporadic CRC, through the evaluation of differential expression of lncRNAs.

The analysis of expression levels of ninety disease-related lncRNAs revealed that

none of the tested transcripts was differently expressed between any pair of groups

compared. Moreover, when comparing all ten CRC samples with the other ten healthy

mucosa samples no significant differences were also found. One explanation for such

lack of significant results may rely on the high inter-individual expression variability of

lncRNAs, even considering the same cell type,162 which suggests not only different


epigenetic patterns in different tissues, but also potential changes environmentally

induced.163 Indeed, most of the selected samples were not paired and the population

studied presented high variability irrespective of clinicopathological variables, including

one patient that received neoadjuvant therapy. Notably, lncRNA annotations differ not

just between tissues, but also between closely related cell types.163 Therefore, a different

location of the studied cases in the large bowel, extracted either from the tumour or

normal mucosa, and a different content of each cell type in those samples, account for

great discrepancies in expression levels. The fact that each lncRNA is often involved in

a wide range of cellular mechanisms with different functions61 is inevitably related with a

higher susceptibility to expression changes upon certain conditions, such as those

occurring during cancer progression; therefore, another major cause of variability may

be attributed to a more or less advanced state of the tumour. In fact, the distribution of

TNM stage among the ten tumour samples is heterogeneous, with three of them being

classified as stage IV. Increasingly, differential expression of lncRNAs has been

associated with tumour TNM stage, mostly higher stages.164

An important aspect to be considered when extracting samples of normal mucosa

from CRC patients is the possible presence of field cancerization, which may extent as

far as 17 cm from the tumour and is initially characterized by sub-cellular alterations,

affecting primarily labile molecular components, such as lncRNAs.165 Thus, the distance

between the tumour and the sample extracted from an apparently normal mucosa may

not be long enough to exclude field cancerization effect, and skewed differences in the

expression levels of lncRNAs.

The small number of representative samples per group coupled with the

existence of many variability factors may have hampered the establishment of compact

relations between any compared pair of formed groups. Although no other work following

the same approach was published so far, many studies exist reporting differently

expressed lncRNAs in CRC. Part of the ninety tested transcripts were formerly reported

as being repressed or induced in CRC.72 Perhaps the most similar study was conducted

by Thorenoor and co-workers (2015), in which tumour and paired non-tumour colorectal

tissues of twenty CRC patients from Czech Republic were screened for the expression

of lncRNAs using the same commercially available qPCR Array Kit. In this independent

work six up-regulated and four down-regulated transcripts were described.166 The only

common transcript differently altered, attending the results depicted herein without

applying any correction to multiple t-tests, was Zfas1. However, contrarily to Thorenoor’s

work, Zfas1 was found to be down-regulated in this present analysis. Moreover, another

work also reported this lncRNA as being up-regulated in CRC and predicting poor

prognosis167. Again, this contradictory observation, and the fact that the majority of the


69

transcripts were suspiciously down-regulated when no correction was applied, is likely

the consequence of great variability and small size of the studied population, further

reflected in the absence of significant results after applying Holm-Šídák correction.

In the future, the analysis performed here should be repeated in a larger

population, and reducing the variability factors to better elucidate the relation between

BER pathway and regulation of lncRNAs in CRC. Moreover, it is important to note that

only ninety transcripts were tested, from a universe of more than ninety thousand

possible lncRNAs genes. High-throughput microarrays represent a more complete

approach, while maintaining statistical power. However, to potentially detect or exclude

any relation at all, the best initial approach would be RNA-sequencing.

The results depicted here suggest that no association exists between BER

pathway and CRC development, considering the expression levels of ninety lncRNAs.

Additionally, none of the transcripts was found to be differently expressed between CRC

tissue and normal colorectal tissue. However, previously published contradictory data,

and high variability and small size of the studied population preclude any solid

conclusion.

PROJECT II

Methylation represents the most well known cancer-related epigenetic alteration.

Because DNA methylation begins early in CRC development, it is the only epigenetic

evidence retained in purified genomic DNA isolated from tumours, and is chemically and

biologically stable. In fact, aberrant DNA methylation is well known to play an important

role not only in cancer onset but also during its progression, and CRC is no exception168

In the past fifteen years, promoter CpG island DNA hypermethylation leading to

transcriptional gene silencing has been recognized as a functional alternative to genetic

mutations inactivating tumour suppressor genes in carcinogenesis. Furthermore, it

should be recalled that CIMP status has been pointed out as the most promising indicator

for prognosticating CRC patients.115 Although CIMP is now collectively accepted as a

subtype of CRCs characterized by epigenetic instability, the same does not applies when

selecting the best approach and group of loci used to define CIMP status of a tumour.

Therefore, in this second project the main goal was the characterization of CIMP status

by specific qMSP in a group of CRC patients using one of the most commonly used

panels, the classic CIMP panel (defined by five markers).

The analysis of typical clinicopathologic and molecular variables distribution for

all CRC patients revealed discrepancies from previously published data. Indeed, in our

series a higher percentage of males was enrolled, when only a slight difference favouring


male sex was reported in larger studied populations. Herein, patients were diagnosed at

earlier age (one-decade difference for median age), which may be the consequence of

a higher percentage of rectal cancers also reported, comparatively to colon tumours.

According to the literature, patients with rectal cancer tend to be younger at diagnosis

than those with colon cancer (median age, 63 vs 70 years, respectively). In this study,

almost half of all patients harboured rectal tumours, in opposition to ~30% reported by

other studies that also included consecutive series. The explanation may be in part

related to lesions missed by colonoscopy, which are more frequently located on the right

side of the bowel, due to poor bowel preparation that prevents complete examination.

Moreover, symptoms are usually easier to notice when the tumour is located on the left

side of the bowel.7,169,170 Nonetheless, the frequency of KRAS mutations are in

accordance with the literature.27 Similarly to others, the majority of tumours were

moderately differentiated (G2), but the frequency of G2 tumours was higher than

previously reported, as almost all tumours were moderately differentiated.169,170 The

percentage of tumours diagnosed with AJCC stage IV was much higher in this study

compared to larger series.169,170 In fact, half of the cases were stage IV, which might be

due to the lack of a CRC screening program in our country and region.

Concerning treatment approaches, the majority of stages II and III tumours was

submitted to neoadjuvant therapy, including mainly rectal tumours. Moreover, half of all

stage IV tumours were also submitted to neadjuvant treatment, most probably

representing unresectable/difficult to resect lesions. Most patients enrolled were

submitted to adjuvant therapy after primary treatment, because 2/3 of all tumours were

classified as stage III or IV. Additionally, adjuvant therapy was applied after progression

of some stage I and II cases.

At some extend these described disparities may also arise due to different follow-

up time considered, different health policies between countries, and diverse inherent

characteristics for different populations, even comparing western developed countries.

However, no similar study was conducted in the Portuguese population, precluding a

better clarification of the subject.

Regarding aberrant methylation, MINT loci and CDKN2A(p16) displayed a lower

proportion (roughly half) of methylated cases for each individual marker than the 20-30%

reported by several authors, consequently affecting CIMP+ frequency.126,171–178

Nevertheless, some studies also reported lower MINT1 and CDKN2A(p16) methylation

frequencies.173–175,178 One plausible explanation for these results would be a high

proportion of rectal cancers, known to be less frequently observed in CIMP+ than colon

tumours.119 However, in this analysis no association was found between tumour location

and CIMP status. Importantly, CIMP+ tumours displaying MLH1 methylation and MSI


71

rarely progress to an advanced stage.178 Indeed, this finding is consistent with the high

proportion of stage IV cases included, the frequency of MLH1 methylation was only

0.95%, in opposition to ~15% usually reported in sporadic colorectal cancers.

Nevertheless, many studies analysing CIMP classic panel also reported methylation

frequencies for MLH1 substantially lower than the other four markers.126,171–173,177–179

Moreover, differences in methodologies may explain some discrepancies. Specifically,

the selected pair of primers and corresponding region of the CpG island in MLH1

promoter may have underrepresented promoter methylation in this gene, as they were

newly designed. In fact, the location of core regions and the density of methylation

required for gene silencing can vary per gene. Thus, to overcome this limitation, a

different pair of primers comprising a more representative region/amplicon may be used

instead.

Similarly, the low frequency of CIMP cases might be due to the quantitative

technique (SYBR® Green-based real-time qMSP) herein performed, since it excludes

most false-positive cases, leading to lower levels of methylation comparatively to non-

quantitative methodology, such as MSP. Moreover, most studies using qMSP prefer

MethyLight as the specific quantitative technique.118,180,181 Since no CIMP studies using

SYBR® Green-based real-time qMSP were found, direct comparisons can not be

performed.

Importantly, one meta-analysis including 33 studies in which CIMP was evaluated

in CRC described a median prevalence of CIMP-positive or CIMP-high status amongst

included studies of 18.2%, ranging from 4.6% to 46.5%.118 Therefore, the frequency of

CIMP+ cases in our study is within the range reported. However, the same is not true

when evaluating a trichotomized categorization of CIMP classic panel – only three cases

were found to be CIMP-H. Nevertheless, the majority of studies testing the classic panel

prefer the dichotomous categorization, which may be related with the specific markers

constituting the panel. Additionally, it was suggested that either a two panel method using

two different sets of CIMP-related markers or an eight-gene panel (such as Ogino’s

panel) are required to properly classify CRC into one of three DNA methylation

epigenotypes.181 Nevertheless, the indicative analysis of frequency for the three cases

was in accordance with the literature, regarding female gender, older age and non-rectal

location.118,120

Interestingly, differences in ethnicity may as well explain why the prevalence of

CIMP differs between study populations, even if the same gene panel and analytic

methods were used in each. Indeed, English et al (2008) found that southern European

origin individuals had lower risk of CRC CIMP+ than people of Anglo-Celtic origin,


possibly owing to genetic factors that are less common in people of southern European

origin.182

In this study no significant associations were apparent between CIMP status and

any of the other molecular and clinicopathological variables. However, according to

published results, CIMP-L and MINT31 methylation were significantly associated with

mutated KRAS. Specifically, KRAS mutation has been associated with CIMP-L tumours,

whereas CIMP+ tumours are associated with BRAF mutation.46,118,120,127 Yet, since part

of CIMP-L tumours are also CIMP+, depending on the proportion of methylated markers

and the threshold used for CIMP panel definition, a trend or even a significant association

between KRAS and CIMP+ tumours may also be reported.104,127 Herein, tumours with

two or three methylated markers were classified as both CIMP-L and CIMP+, which

probably accounts for the trend observed for CIMP+ tumours to be KRAS mutated.

Nonetheless, it has been suggested that BRAF and KRAS oncogene mutation

status may refine CIMP definition.118 Therefore, an important complement to this study

would be the screening for BRAF mutational status.

Surprisingly, CIMP-L significantly associated with a higher number of regional

lymph nodes with metastasis (N2), relatively to N1 tumours. Similarly, MINT1 methylation

significantly associated with lack of lymph nodes metastasis (N0). However, due to the

low number of tumours with methylated MINT1 promoter, N1 and N2 cases were

combined in the same category, which precluded the comparison between N1 and N2

cases and may have negatively affected the statistical value of the test for MINT1 locus.

These differences are most probably due to the population analysed, since lymph node

metastasis (N) status was not previously associated with CIMP-L or the methylation of

this MINT loci. Moreover, CIMP-L tumours, comparatively to CIMP-H, are not as

commonly correlated with poor prognosis120, and thus, no association with higher N stage

is expected.

Concerning survival analysis, several clinicopathological parameters have been

previously described as being associated with CRC prognosis, including age, gender,

tumour grade, depth of tumour growth, lymph node metastasis, distant metastasis and

staging.183 Indeed, in our study, all variables but gender associated with shorter DSS.

Moreover, neoadjuvant therapy was also independently associated with shorter DSS,

which is related with poorer outcome of unresectable metastatic tumours. Still a great

debate exists about the eventual outcome improvement of neoadjuvant therapy in these

patients.184,185 Regarding DFS, only male gender was independently associated with

poorer prognosis. These discrepancies may be at least in part explained by the

substantial reduction of cases considered for DFS analysis.


73

It should be recalled that our cohort of patients includes a high proportion of

tumours with advanced stages of CRC (metastatic mostly), thus impacting both in DSS

and DFS, which is in line with a poorer prognosis depicted for stage IV tumours, in

DSS.7,186

Although CIMP is generally accepted as a predictor of worse prognosis, in the

present study CIMP status showed no prognostic value both for DSS and DFS. However,

these results follow the same trend as in a recent meta-analysis in which 13 out of 19

studies concluded that CIMP had no significant effect on OS, and 8 of 11 studies found

no significant relationship between CIMP and DFS.118 Moreover, in another meta-

analysis all four studies reporting the effect of CIMP in DSS, without considering any

subgroup of patients, found no significant association between CIMP+ tumours and

survival; and three out of four studies considering exclusively the classic panel showed

no significant association.180 Therefore, the prognostic value of CIMP may not only

depend on the specific population studied and the associated characteristics, but also

according to the panel selected.

Importantly, CIMP tumours have been strongly associated with worse outcome

when considering only MSS and MSI-L tumours.118 However, MSI profiling was not

performed, and therefore, it is not possible to test whether MSI status would alter the

results, which limits the potential of this study.

Of notice, all three CIMP-H patients died from disease progression, and two of

them were never considered cured during the follow-up time, which is in line with the

frequently reported association between CIMP-H and worse prognosis.120

From all individual markers, only CDKN2A(p16) aberrant methylation significantly

associated with poor prognosis in DSS and univariable analysis. However, significance

was lost after multivariable analysis, which is in agreement with a large cohort study

examining the prognostic effect of this gene promoter methylation independent of

CIMP.187 Nevertheless, the prognostic significance of CDKN2A(p16) methylation

independent of CIMP status remains uncertain. Specifically, a recent meta-analysis

suggests that CDKN2A(p16) methylation might be a predictive factor for unfavourable

prognosis of CRC patients.188 In line with this, after analysing only those cases not

submitted to neoadjuvant therapy, an independent association between CDKN2A(p16)

methylation and worse prognosis was found. Even though CDKN2A(p16) methylation is

often included in the CIMP panel and is closely related to CIMP status, the reported age-

related CDKN2A(p16) methylation likely represents a confounding factor in the

assessment of tumour-specific methylation and subsequent correlation with

outcome.117,187 Nevertheless, as mentioned above, no correlation between


CDKN2A(p16) promoter methylation and age at diagnosis was found in the present

study.

Currently, although KRAS mutations are acknowledged as a predictive marker in

anti-EGFR therapy, its value as a prognostic marker is highly questionable.189 In this

dissertation, KRAS mutations did not associate with survival. However, KRAS mutations

predicted worse prognosis in MINT31 methylated cases, whereas in CIMP-L tumours

associated with better outcome. Nonetheless, MINT31 methylation and CIMP-L were

both associated with improved outcome in KRAS mutated tumours. In fact, CIMP-L was

significantly associated with contradictory outcomes in KRAS wild-type and mutated

tumours. These intricate results may be related with the association between MINT31 or

CIMP-L with mutation of KRAS found in our cohort of patients. However, none of these

or other significant associations observed after stratification was described in previous

studies. Therefore, to further clarify or validate these new findings and its possible

implications in CRC prognostication, a new and independent series should be analysed.

Additionally, numerous gene panel definitions, and different marker thresholds

and laboratory methods have been used to study CIMP in CRC, which has been shown

to result in varied CIMP frequencies and different conclusions regarding the prognostic

value of CIMP. This lack of consensus is surely related to the still unknown biological

cause of CIMP tumours. In addition, a difference in the choice of primers and/or the

precise location of the region analysed to determine methylation of the marker may as

well explain discrepancies observed between studies. Therefore, in order to further

determine the relation between CIMP status and survival or treatment response, the

eventual effect of MSI, BRAF, and KRAS status should be taken into consideration.

The fact that qMSP (MethyLight) has been most frequently used alongside with

Weisenberger’s (new) panel may point out a brittleness of the present analysis. However,

using qMSP alongside with the classic panel instead, in fact, adds more information to

the discussion of which is the best approach to profile CIMP in CRC. To complement this

analysis, MSI and BRAF status should be evaluated. Analysing the same population

following an identical approach but with a different panel or method would be of great

importance, allowing for a more direct comparison of the potential of each panel or the

feasibility of each laboratory technique.

In conclusion, the analysis of CIMP status in this set of 211 CRC patients

revealed that CIMP+ phenotype is rare in sporadic CRC and does not have an

independent prognostic value in this malignancy.


75

REFERENCES 1. Slyskova, J. et al. Functional, genetic, and epigenetic aspects of base and nucleotide excision repair

in colorectal carcinomas. Clin. Cancer Res. 18, 5878–87 (2012). 2. World Health Organization. World Cancer Report 2014. (World Health Organization, 2014). 3. Schreuders, E. H. et al. Colorectal cancer screening: a global overview of existing programmes. Gut

1–13 (2015). doi:10.1136/gutjnl-2014-309086 4. Ait Ouakrim, D. et al. Trends in colorectal cancer mortality in Europe: retrospective analysis of the

WHO mortality database. BMJ 351, h4970 (2015). 5. Murphy, G. et al. Sex disparities in colorectal cancer incidence by anatomic subsite, race and age.

Int. J. Cancer 128, 1668–1675 (2011). 6. American Cancer Society. Colorectal Cancer Facts & Figures 2014-2016. (2014).

doi:10.1101/gad.1593107 7. DeSantis, C. E. et al. Cancer Treatment and Survivorship Statistics, 2014. CA. Cancer J. Clin. 64,

252–271 (2014). 8. Siegel, R., Naishadham, D. & Jemal, A. Cancer statistics, 2013. CA. Cancer J. Clin. 63, 11–30

(2013). 9. Inra, J. A. & Syngal, S. Colorectal Cancer in Young Adults. Dig. Dis. Sci. 60, 722–733 (2015). 10. Coppedè, F. Epigenetic biomarkers of colorectal cancer: Focus on DNA methylation. Cancer Letters

342, 238–247 (2014). 11. Ogino, S., Chan, A. T., Fuchs, C. S. & Giovannucci, E. Molecular pathological epidemiology of

colorectal neoplasia: an emerging transdisciplinary and interdisciplinary field. Gut 60, 397–411 (2011).

12. Hol, L. et al. Screening for colorectal cancer: random comparison of guaiac and immunochemical faecal occult blood testing at different cut-off levels. Br. J. Cancer 100, 1103–1110 (2009).

13. Budinska, E. et al. Gene expression patterns unveil a new level of molecular heterogeneity in colorectal cancer. J. Pathol. 231, 63–76 (2013).

14. Kahn, E. D. F. in Sleisenger and Fordtran’s Gastrointestinal and Liver Disease - 2 Volume Set (ed. Mark Feldman, Lawrence S. Friedman, L. J. B.) 1615–1640 (Saunders, 2010). doi:10.1016/B978-1-4160-6189-2.00096-2

15. Li, F. & Lai, M. Colorectal cancer, one entity or three. J. Zhejiang Univ. Sci. B 10, 219–229 (2009). 16. World Health Organization. International Classification of Diseases for Oncology (ICD-O). (World

Health Organization, 2000). doi:10.1136/jcp.30.8.782-c 17. Schuebel, K., Chen, W. & Baylin, S. B. CIMPle origin for promoter hypermethylation in colorectal

cancer? Nat Genet 38, 738–740 (2006). 18. Stryker, S. J. et al. Natural history of untreated colonic polyps. Gastroenterology 93, 1009–13 (1987). 19. Carethers, J. M. & Jung, B. H. Genetics and Genetic Biomarkers in Sporadic Colorectal Cancer.

Gastroenterology 149, 1177–1190 (2015). 20. Damiens, K. et al. Clinical features and course of brain metastases in colorectal cancer: an

experience from a single institution. Curr. Oncol. 19, 254–8 (2012). 21. Buetow, P. C., Buck, J. L., Carr, N. J. & Pantongrag-Brown, L. From the archives of the AFIP.

Colorectal adenocarcinoma: radiologic-pathologic correlation. RadioGraphics 15, 127–146 (1995). 22. Karim Alwan Al-Jashamy. in Principles and practice of Cancer prevention and control (ed. Al-

Naggar, R. A.) (OMICS Group International, 2014). 23. Hisamuddin, I. M. & Yang, V. W. Molecular Genetics of Colorectal Cancer: An Overview. Curr Color.

Cancer Rep 2, 53–59 (2006). 24. Hisamuddin, I. M. & Yang, V. W. Genetics of Colorectal Cancer. Medscape Gen. Med. 6, 13 (2004). 25. The Cancer Genome Atlas Network. Comprehensive molecular characterization of human colon and

rectal cancer. Nature 487, 330–7 (2012). 26. Fearon, E. R. & Vogelstein, B. A genetic model for colorectal tumorigenesis. Cell 61, 759–767

(1990). 27. Markowitz, S. D. & Bertagnolli, M. M. Molecular basis of colorectal cancer. N. Engl. J. Med. 362,

1246; author reply 1246–1247 (2009). 28. Vogelstein, B. et al. Cancer Genome Landscapes. Science (80-. ). 339, 1546–1558 (2013). 29. Pino, M. S. & Chung, D. C. The Chromosomal Instability Pathway in Colon Cancer. Gastroenterology

138, 2059–2072 (2010). 30. Choong, M. K. & Tsafnat, G. Genetic and epigenetic biomarkers of colorectal cancer. Clinical

Gastroenterology and Hepatology 10, 9–15 (2012). 31. Labianca, R. et al. Early colon cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment

and follow-up. Ann. Oncol. 24 , vi64–vi72 (2013). 32. Van Cutsem, E., Cervantes, A., Nordlinger, B. & Arnold, D. Metastatic colorectal cancer: ESMO

Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann. Oncol. 25 , iii1–iii9 (2014). 33. Glimelius, B., Tiret, E., Cervantes, A., Arnold, D. & Group, on behalf of the E. G. W. Rectal cancer:

ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann. Oncol. 24 , vi81–vi88 (2013).

34. Sobin, L. H., Gospodarowicz, M. K. & Wittekind, C. TNM classification of malignant tumours. Clinical


Oncology 10, (2009).

35. Tsai, H.-C. & Baylin, S. B. Cancer epigenetics: linking basic biology to clinical medicine. Cell Res. 21, 502–517 (2011).

36. Lao, V. V. & Grady, W. M. Epigenetics and colorectal cancer. Nat. Rev. Gastroenterol. Hepatol. 8, 686–700 (2011).

37. Waddington, C. H. Preliminary Notes on the Development of the Wings in Normal and Mutant Strains of Drosophila. Proc. Natl. Acad. Sci. U. S. A. 25, 299–307 (1939).

38. Feinberg, A. P. & Vogelstein, B. Hypomethylation distinguishes genes of some human cancers from their normal counterparts. Nature 301, 89–92 (1983).

39. Van Engeland, M., Derks, S., Smits, K. M., Meijer, G. A. & Herman, J. G. Colorectal cancer epigenetics: Complex simplicity. J. Clin. Oncol. 29, 1382–1391 (2011).

40. You, J. S. & Jones, P. A. Cancer Genetics and Epigenetics: Two Sides of the Same Coin? Cancer Cell 22, 9–20 (2012).

41. Luger, K., Dechassa, M. L. & Tremethick, D. J. New insights into nucleosome and chromatin structure: an ordered state or a disordered affair? Nat. Rev. Mol. Cell Biol. 13, 436–447 (2012).

42. Vaiopoulos, A. G., Kostakis, I. D., Athanasoula, K. C. & Papavassiliou, A. G. Targeting transcription factor corepressors in tumor cells. Cell. Mol. Life Sci. 69, 1745–1753 (2012).

43. Narlikar, G. J., Sundaramoorthy, R. & Owen-Hughes, T. Mechanisms and functions of ATP-dependent chromatin-remodeling enzymes. Cell 154, 490–503 (2013).

44. Dawson, M. A. & Kouzarides, T. Cancer epigenetics: From mechanism to therapy. Cell 150, 12–27 (2012).

45. Vaiopoulos, A. G., Athanasoula, K. C. & Papavassiliou, A. G. Epigenetic modifications in colorectal cancer: molecular insights and therapeutic challenges. Biochim. Biophys. Acta 1842, 971–80 (2014).

46. Coppedè, F. The role of epigenetics in colorectal cancer. Expert Rev. Gastroenterol. Hepatol. 8, 935–948 (2014).

47. Stypula-Cyrus, Y. et al. HDAC Up-Regulation in Early Colon Field Carcinogenesis Is Involved in Cell Tumorigenicity through Regulation of Chromatin Structure. PLoS One 8, (2013).

48. Rönsch, K. et al. Class I and III HDACs and loss of active chromatin features contribute to epigenetic silencing of CDX1 and EPHB tumor suppressor genes in colorectal cancer. Epigenetics 6, 610–622 (2011).

49. Jie, D. et al. Positive Expression of LSD1 and Negative Expression of E-cadherin Correlate with Metastasis and Poor Prognosis of Colon Cancer. Dig. Dis. Sci. 58, 1581–1589 (2013).

50. Simon, J. a & Kingston, R. E. Mechanisms of polycomb gene silencing: knowns and unknowns. Nat. Rev. Mol. Cell Biol. 10, 697–708 (2009).

51. Djebali, S. et al. Landscape of transcription in human cells. Nature 489, 101–8 (2012). 52. Slaby, O., Svoboda, M., Michalek, J. & Vyzula, R. MicroRNAs in colorectal cancer: translation of

molecular biology into clinical application. Mol. Cancer 8, 102 (2009). 53. Sato, F., Tsuchiya, S., Meltzer, S. J. & Shimizu, K. MicroRNAs and epigenetics. FEBS Journal 278,

1598–1609 (2011). 54. Goel, A. & Boland, C. R. Epigenetics of Colorectal Cancer. Gastroenterology 143, 1442–1460.e1

(2012). 55. Aslam, M. I., Patel, M., Singh, B., Jameson, J. S. & Pringle, J. H. MicroRNA manipulation in

colorectal cancer cells: from laboratory to clinical application. J. Transl. Med. 10, 128 (2012). 56. Chen, X. et al. Constructing lncRNA functional similarity network based on lncRNA-disease

associations and disease semantic similarity. Sci. Rep. 5, 11338 (2015). 57. Gutschner, T. & Diederichs, S. The hallmarks of cancer: a long non-coding RNA point of view. RNA

Biol. 9, 703–19 (2012). 58. Wilusz, J. E., Sunwoo, H. & Spector, D. L. Long noncoding RNAs: functional surprises from the RNA

world. Genes Dev. 23, 1494–504 (2009). 59. Schonrock, N., Harvey, R. P. & Mattick, J. S. Long noncoding RNAs in cardiac development and

pathophysiology. Circulation Research 111, 1349–1362 (2012). 60. Guttman, M. & Rinn, J. L. Modular regulatory principles of large non-coding RNAs. Nature 482, 339–

346 (2012). 61. Rinn, J. L. & Chang, H. Y. Genome regulation by long noncoding RNAs. Annu. Rev. Biochem. 81,

145–166 (2012). 62. Li, C. H. & Chen, Y. Targeting long non-coding RNAs in cancers: Progress and prospects.

International Journal of Biochemistry and Cell Biology 45, 1895–1910 (2013). 63. Han, D. et al. Long noncoding RNAs: Novel players in colorectal cancer. Cancer Lett. 361, 13–21

(2015). 64. Huarte, M. The emerging role of lncRNAs in cancer. Nat Med 21, 1253–1261 (2015). 65. Ye, L. C., Zhu, X., Qiu, J. J., Xu, J. & Wei, Y. Involvement of long non-coding RNA in colorectal

cancer: From benchtop to bedside (Review). Oncol Lett 9, 1039–1045 (2015). 66. Lee, H. et al. A Long Non-Coding RNA snaR Contributes to 5-Fluorouracil Resistance in Human

Colon Cancer Cells. Mol Cells (2014). doi:10.14348/molcells.2014.0151 67. Yu, Z. Q. et al. [Long non-coding RNA influences radiosensitivity of colorectal carcinoma cell lines

by regulating cyclin D1 expression]. Zhonghua Wei Chang Wai Ke Za Zhi 15, 288–291 (2012). 68. Sánchez, Y. et al. Genome-wide analysis of the human p53 transcriptional network unveils a lncRNA


77

tumour suppressor signature. Nat. Commun. 5, 5812 (2014).

69. Kim, T. et al. Role of MYC-Regulated long noncoding RNAs in cell cycle regulation and tumorigenesis. J. Natl. Cancer Inst. 107, (2015).

70. Liao, Q. et al. Identification and functional annotation of lncRNA genes with hypermethylation in colorectal cancer. Gene 572, 259–265 (2015).

71. Nissan, A. et al. Colon cancer associated transcript-1: A novel RNA expressed in malignant and pre-malignant human tissues. Int. J. Cancer 130, 1598–1606 (2012).

72. Xie, X. et al. Long non-coding RNAs in colorectal cancer. Oncotarget 7, (2015). 73. Ling, H. et al. CCAT2, a novel noncoding RNA mapping to 8q24, underlies metastatic progression

and chromosomal instability in colon cancer. Genome Res. 23, 1446–1461 (2013). 74. Xiang, J.-F. et al. Human colorectal cancer-specific CCAT1-L lncRNA regulates long-range

chromatin interactions at the MYC locus. Cell Res. 24, 513–531 (2014). 75. Graham, L. D. et al. Colorectal Neoplasia Differentially Expressed (CRNDE), a Novel Gene with

Elevated Expression in Colorectal Adenomas and Adenocarcinomas. Genes and Cancer 2, 829–840 (2011).

76. Liang, W.-C. et al. The lncRNA H19 promotes epithelial to mesenchymal transition by functioning as miRNA sponges in colorectal cancer. Oncotarget 6, 22513–22525 (2015).

77. Keniry, A. et al. The H19 lincRNA is a developmental reservoir of miR-675 that suppresses growth and Igf1r. Nat. Cell Biol. 14, 659–665 (2012).

78. Xu, M.-D., Qi, P. & Du, X. Long non-coding RNAs in colorectal cancer: implications for pathogenesis and clinical application. Mod. Pathol. 1–11 (2014). doi:10.1038/modpathol.2014.33

79. Kogo, R. et al. Long noncoding RNA HOTAIR regulates polycomb-dependent chromatin modification and is associated with poor prognosis in colorectal cancers. Cancer Res. 71, 6320–6326 (2011).

80. Ji, Q. et al. Long non-coding RNA MALAT1 promotes tumour growth and metastasis in colorectal cancer through binding to SFPQ and releasing oncogene PTBP2 from SFPQ/PTBP2 complex. Br. J. Cancer 111, 736–748 (2014).

81. Xu, C., Yang, M., Tian, J., Wang, X. & Li, Z. MALAT-1: A long non-coding RNA and its important 3??? end functional motif in colorectal cancer metastasis. Int. J. Oncol. 39, 169–175 (2011).

82. Wang, J. et al. CREB up-regulates long non-coding RNA, HULC expression through interaction with microRNA-372 in liver cancer. Nucleic Acids Res. 38, 5366–5383 (2010).

83. Takahashi, Y. et al. Amplification of PVT-1 is involved in poor prognosis via apoptosis inhibition in colorectal cancers. Br. J. Cancer 110, 164–71 (2014).

84. Han, Y. J., Ma, S. F., Yourek, G., Park, Y.-D. & Garcia, J. G. N. A transcribed pseudogene of MYLK promotes cell proliferation. FASEB J. 25, 2305–2312 (2011).

85. Prensner, J. R. et al. PCAT-1, a long noncoding RNA, regulates BRCA2 and controls homologous recombination in cancer. Cancer Res. 74, 1651–1660 (2014).

86. Ge, X. et al. Overexpression of long noncoding RNA PCAT-1 is a novel biomarker of poor prognosis in patients with colorectal cancer. Med. Oncol. 30, 588 (2013).

87. Zhou, Y. et al. Activation of p53 by MEG3 non-coding RNA. J. Biol. Chem. 282, 24731–24742 (2007).

88. Yang, F. et al. Repression of the Long Noncoding RNA-LET by Histone Deacetylase 3 Contributes to Hypoxia-Mediated Metastasis. Mol. Cell 49, 1083–1096 (2016).

89. Liu, Q. et al. LncRNA loc285194 is a p53-regulated tumor suppressor. Nucleic Acids Res. 41, 4976–4987 (2013).

90. Zhai, H. et al. Clinical significance of long intergenic noncoding RNA-p21 in colorectal Cancer. Clin. Colorectal Cancer 12, 261–266 (2013).

91. Poliseno, L. et al. A coding-independent function of gene and pseudogene mRNAs regulates tumour biology. Nature 465, 1033–8 (2010).

92. Yin, D. et al. Long noncoding RNA GAS5 affects cell proliferation and predicts a poor prognosis in patients with colorectal cancer. Med. Oncol. 31, 253 (2014).

93. Jackson, S. P. & Bartek, J. The DNA-damage response in human biology and disease. Nature 461, 1071–1078 (2010).

94. Ciccia, A. & Elledge, S. J. The DNA Damage Response: Making It Safe to Play with Knives. Mol. Cell 40, 179–204 (2010).

95. Khanna, K. K. & Jackson, S. P. DNA double-strand breaks: signaling, repair and the cancer connection. Nat. Genet. 27, 247–254 (2001).

96. Wang, L. et al. BRCA1 is a negative modulator of the PRC2 complex. EMBO J. 32, 1584–1597 (2013).

97. Wan, G. et al. A novel non-coding RNA lncRNA-JADE connects DNA damage signalling to histone H4 acetylation. EMBO J. 32, 2833–47 (2013).

98. Sharma, V. et al. A BRCA1-interacting lncRNA regulates homologous recombination. EMBO Rep. 16, 1520–1534 (2015).

99. Gazy, I. et al. TODRA, a lncRNA at the RAD51 locus, is oppositely regulated to RAD51, and enhances RAD51-dependent DSB (double strand break) repair. PLoS One 10, (2015).

100. Wan, G. et al. Long non-coding RNA ANRIL (CDKN2B-AS) is induced by the ATM-E2F1 signaling pathway. Cell. Signal. 25, 1086–1095 (2013).


101. Zhang, H. et al. Targeting WISP1 to sensitize esophageal squamous cell carcinoma to irradiation.

Oncotarget 6, 6218–6234 (2015). 102. Li, H. et al. LncRNA HOTAIR promotes human liver cancer stem cell malignant growth through

downregulation of SETD2. Oncotarget 6, (2014). 103. Nissar, S., Sameer, A. S., Rasool, R. & Rashid, F. DNA repair gene--XRCC1 in relation to genome

instability and role in colorectal carcinogenesis. Oncol. Res. Treat. 37, 418–22 (2014). 104. Mojarad, E. N., Kuppen, P. J. K., Aghdaei, H. A. & Zali, M. R. The CpG island methylator phenotype

(CIMP) in colorectal cancer. Gastroenterology and Hepatology from Bed to Bench 6, 120–128 (2013).

105. Lee, J. K. & Chan, A. T. Molecular prognostic and predictive markers in colorectal cancer: Current status. Current Colorectal Cancer Reports 7, 136–144 (2011).

106. Pawlik, T. M., Raut, C. P. & Rodriguez-Bigas, M. A. Colorectal carcinogenesis: MSI-H versus MSI-L. Dis. Markers 20, 199–206 (2004).

107. Howard, J. H. et al. Epigenetic downregulation of the DNA repair gene MED1/MBD4 in colorectal and ovarian cancer. Cancer Biol. Ther. 8, 1–7 (2009).

108. Shen, L. et al. MGMT promoter methylation and field defect in sporadic colorectal cancer. J. Natl. Cancer Inst. 97, 1330–1338 (2005).

109. Farkas, S. A., Vymetalkova, V., Vodickova, L., Vodicka, P. & Nilsson, T. K. DNA methylation changes in genes frequently mutated in sporadic colorectal cancer and in the DNA repair and Wnt/β-catenin signaling pathway genes. Epigenomics 6, 179–191 (2014).

110. Santos, J. C. et al. Effect of APE1 T2197G (Asp148Glu) Polymorphism on APE1, XRCC1, PARP1 and OGG1 Expression in Patients with Colorectal Cancer. Int. J. Mol. Sci. 15, 17333–17343 (2014).

111. Kabzinski, J. et al. Efficiency of Base Excision Repair of Oxidative DNA Damage and Its Impact on the Risk of Colorectal Cancer in the Polish Population. Oxid. Med. Cell. Longev. 2016, (2015).

112. Wood, R. D., Mitchell, M., Sgouros, J. & Lindahl, T. Human DNA repair genes. Science 291, 1284–9 (2001).

113. Fuss, J. O. & Cooper, P. K. DNA repair: Dynamic defenders against cancer and aging. PLoS Biol. 4, 0899–0903 (2006).

114. Wyatt, M. D. & Wilson, D. M. Participation of DNA repair in the response to 5-fluorouracil. Cellular and Molecular Life Sciences 66, 788–799 (2009).

115. Okugawa, Y., Grady, W. M. & Goel, A. Epigenetic Alterations in Colorectal Cancer: Emerging Biomarkers. Gastroenterology 149, 1204–1225e12 (2015).

116. Kim, M. S., Lee, J. & Sidransky, D. DNA methylation markers in colorectal cancer. Cancer and Metastasis Reviews 29, 181–206 (2010).

117. Toyota, M. et al. CpG island methylator phenotype in colorectal cancer. Proc. Natl. Acad. Sci. U. S. A. 96, 8681–8686 (1999).

118. Juo, Y. Y. et al. Prognostic value of CpG island methylator phenotype among colorectal cancer patients: a systematic review and meta-analysis. Ann. Oncol. 25, 2314–27 (2014).

119. Yamauchi, M. et al. Assessment of colorectal cancer molecular features along bowel subsites challenges the conception of distinct dichotomy of proximal versus distal colorectum. Gut 61, 847–54 (2012).

120. Ogino, S. & Goel, A. Molecular Classification and Correlates in Colorectal Cancer. J. Mol. Diagn. 10, 13–27 (2008).

121. Curtin, K., Slattery, M. L. & Samowitz, W. S. CpG island methylation in colorectal cancer: past, present and future. Patholog. Res. Int. 2011, 902674 (2011).

122. Issa, J.-P. J. Methylation and Prognosis. Clin. Cancer Res. 9, 2879 LP – 2881 (2003). 123. Collado, M., Blasco, M. A. & Serrano, M. Cellular Senescence in Cancer and Aging. Cell 130, 223–

233 (2007). 124. Park, S.-J. et al. Frequent CpG island methylation in serrated adenomas of the colorectum. Am. J.

Pathol. 162, 815–822 (2003). 125. Weisenberger, D. J. et al. CpG island methylator phenotype underlies sporadic microsatellite

instability and is tightly associated with BRAF mutation in colorectal cancer. Nat. Genet. 38, 787–93 (2006).

126. Barault, L. et al. Hypermethylator phenotype in sporadic colon cancer: Study on a population-based series of 582 cases. Cancer Res. 68, 8541–8546 (2008).

127. Ogino, S., Kawasaki, T., Kirkner, G. J., Loda, M. & Fuchs, C. S. CpG island methylator phenotype-low (CIMP-low) in colorectal cancer: possible associations with male sex and KRAS mutations. J. Mol. Diagn. 8, 582–8 (2006).

128. Ogino, S. et al. Evaluation of Markers for CpG Island Methylator Phenotype (CIMP) in Colorectal Cancer by a Large Population-Based Sample. J. Mol. Diagn. 9, 305–314 (2007).

129. Issa, J.-P. Colon cancer: it’s CIN or CIMP. Clin. Cancer Res. 14, 5939–40 (2008). 130. Fleming, M., Ravula, S., Tatishchev, S. F. & Wang, H. L. Colorectal carcinoma: Pathologic aspects.

J. Gastrointest. Oncol. 3, 153–73 (2012). 131. Cheng, Y. W. et al. CpG island methylator phenotype associates with low-degree chromosomal

abnormalities in colorectal cancer. Clin. Cancer Res. 14, 6005–6013 (2008). 132. Sinicrope, F. A. et al. Prognostic impact of microsatellite instability and DNA ploidy in human colon

carcinoma patients. Gastroenterology 131, 729–737 (2006).


79

133. Shen, L. et al. Integrated genetic and epigenetic analysis identifies three different subclasses of

colon cancer. Proc. Natl. Acad. Sci. U. S. A. 104, 18654–9 (2007). 134. Hattori, N. & Ushijima, T. in Handbook of Epigenetics 125–134 (2011). doi:10.1016/B978-0-12-

375709-8.00008-3 135. Olkhov-Mitsel, E. & Bapat, B. Strategies for discovery and validation of methylated and

hydroxymethylated DNA biomarkers. Cancer Med. 1, 237–60 (2012). 136. Atkinson A.J., J. et al. Biomarkers and surrogate endpoints: Preferred definitions and conceptual

framework. Clinical Pharmacology and Therapeutics 69, 89–95 (2001). 137. Coppedè, F., Lopomo, A., Spisni, R. & Migliore, L. Genetic and epigenetic biomarkers for diagnosis,

prognosis and treatment of colorectal cancer. World J. Gastroenterol. 20, 943–56 (2014). 138. Gyparaki, M. T., Basdra, E. K. & Papavassiliou, A. G. DNA methylation biomarkers as diagnostic

and prognostic tools in colorectal cancer. Journal of Molecular Medicine 91, 1249–1256 (2013). 139. Imperiale, T. F. et al. Multitarget stool DNA testing for colorectal-cancer screening. N. Engl. J. Med.

370, 1287–97 (2014). 140. Gonzalez-pons, M. & Cruz-correa, M. Colorectal Cancer Biomarkers : Where Are We Now ? Biomed

Res. Int. 2015, 1–14 (2015). 141. Jenab-Wolcott, J. & Giantonio, B. in Molecular Pathology of Neoplastic Gastrointestinal Diseases

(eds. Sepulveda, R. A. & Lynch, P. J.) 141–171 (Springer US, 2013). doi:10.1007/978-1-4614-6015-2_9

142. Nilsson, T. K., Lof-Ohlin, Z. M. & Sun, X.-F. DNA methylation of the p14(ARF), RASSF1A and APC1A genes as an independent prognostic factor in colorectal cancer patients. Int. J. Oncol. 42, 127–133 (2013).

143. Wallner, M. et al. Methylation of serum DNA is an independent prognostic marker in colorectal cancer. Clin. Cancer Res. 12, 7347–7352 (2006).

144. Jensen, L. H. et al. Regulation of MLH1 mRNA and protein expression by promoter methylation in primary colorectal cancer: A descriptive and prognostic cancer marker study. Cell. Oncol. 36, 411–419 (2013).

145. Sharma, A., Abdelfatah, E., Al Eissa, M. & Ahuja, N. in Medical Epigenetics (ed. Tollefsbol, T. O.) 177–195 (Academic Press, 2016). doi:http://dx.doi.org/10.1016/B978-0-12-803239-8.00011-9

146. Cha, Y. et al. Adverse prognostic impact of the CpG island methylator phenotype in metastatic colorectal cancer. Br J Cancer 115, 164–171 (2016).

147. Joo, Y.-E. Prognostic Significance of CpG Island Methylator Phenotype in Colorectal Cancer. Gut Liver 9, 139–140 (2015).

148. Bae, J. M., Kim, J. H., Cho, N.-Y., Kim, T.-Y. & Kang, G. H. Prognostic implication of the CpG island methylator phenotype in colorectal cancers depends on tumour location. Br. J. Cancer 109, 1004–12 (2013).

149. Samowitz, W. S. et al. Association of smoking, CpG island methylator phenotype, and V600E BRAF mutations in colon cancer. J. Natl. Cancer Inst. 98, 1731–1738 (2006).

150. Murcia, O. et al. Serrated colorectal cancer: Molecular classification, prognosis, and response to chemotherapy. World Journal of Gastroenterology 22, 3516–3530 (2016).

151. Langie, S. A. S. et al. Measuring DNA repair incision activity of mouse tissue extracts towards singlet oxygen-induced DNA damage: A comet-based in vitro repair assay. Mutagenesis 26, 461–471 (2011).

152. Shaposhnikov, S. et al. Twelve-gel slide format optimised for comet assay and fluorescent in situ hybridisation. Toxicol. Lett. 195, 31–34 (2010).

153. Olive, P. L. & Banáth, J. P. The comet assay: a method to measure DNA damage in individual cells. Nat. Protoc. 1, 23–29 (2006).

154. Ririe, K. M., Rasmussen, R. P. & Wittwer, C. T. Product Differentiation by Analysis of DNA Melting Curves during the Polymerase Chain Reaction. Anal. Biochem. 245, 154–160 (1997).

155. Hamilton, S. R., Aaltonen, L. a, Kleihues, P. & Cavenee, W. K. World health organization. Classification of tumours. Pathology and genetics of tumours of the digestive system. Parallel Distrib. Process. 2008 IPDPS 2008 IEEE Int. Symp. 1–8 (2000).

156. Egner, J. R. AJCC Cancer Staging Manual. JAMA: The Journal of the American Medical Association 304, 1726 (2010).

157. Pearson, H. & Stirling, D. in PCR Protocols (eds. Bartlett, J. M. S. & Stirling, D.) 33–34 (Humana Press, 2003). doi:10.1385/1-59259-384-4:33

158. Frommer, M. et al. A genomic sequencing protocol that yields a positive display of 5-methylcytosine residues in individual DNA strands. Proc. Natl. Acad. Sci. U. S. A. 89, 1827–31 (1992).

159. Li, Y. & Tollefsbol, T. O. DNA methylation detection: Bisulfite genomic sequencing analysis. Methods Mol. Biol. 791, 11–21 (2011).

160. Toyota, M. et al. Identification of differentially methylated sequences in colorectal cancer by methylated CpG island amplification. Cancer Res. 59, 2307–2312 (1999).

161. Kondo, Y., Shen, L. & Issa, J.-P. J. Critical role of histone methylation in tumor suppressor gene silencing in colorectal cancer. Mol. Cell. Biol. 23, 206–15 (2003).

162. Kornienko, A. E. et al. Long non-coding RNAs display higher natural expression variation than protein-coding genes in healthy humans. Genome Biol. 17, 14 (2016).

163. Amin, V. et al. Epigenomic footprints across 111 reference epigenomes reveal tissue-specific


epigenetic regulation of lincRNAs. Nat. Commun. 6, 6370 (2015).

164. Kladi-Skandali, A., Michaelidou, K., Scorilas, A. & Mavridis, K. Long Noncoding RNAs in Digestive System Malignancies: A Novel Class of Cancer Biomarkers and Therapeutic Targets? Gastroenterol. Res. Pract. 2015, 319861 (2015).

165. Patel, A., Tripathi, G., Gopalakrishnan, K., Williams, N. & Arasaradnam, R. P. Field cancerisation in colorectal cancer: A new frontier or pastures past? World Journal of Gastroenterology 21, 3763–3772 (2015).

166. Thorenoor, N. et al. Long non-coding RNA ZFAS1 interacts with CDK1 and is involved in p53-dependent cell cycle control and apoptosis in colorectal cancer. Oncotarget 7, 622–637 (2016).

167. Wang, W. & Xing, C. Upregulation of long noncoding RNA ZFAS1 predicts poor prognosis and prompts invasion and metastasis in colorectal cancer. Pathol. - Res. Pract. 212, 690–695 (2016).

168. Crider, K. S., Yang, T. P., Berry, R. J. & Bailey, L. B. Folate and DNA Methylation : A Review of Molecular Mechanisms and the Evidence for Folate ’ s Role. Am. Soc. Nutr. 3, 21–38 (2012).

169. Erichsen, R. et al. Characteristics and survival of interval and sporadic colorectal cancer patients: a nationwide population-based cohort study. Am J Gastroenterol 108, 1332–1340 (2013).

170. Derwinger, K., Kodeda, K., Bexe-Lindskog, E. & Taflin, H. Tumour differentiation grade is associated with TNM staging and the risk of node metastasis in colorectal cancer. Acta Oncol. (Madr). 49, 57–62 (2010).

171. Dahlin, A. M. et al. The role of the CpG island methylator phenotype in colorectal cancer prognosis depends on microsatellite instability screening status. Clin. Cancer Res. 16, 1845–1855 (2010).

172. Han, S. W. et al. Methylation and microsatellite status and recurrence following adjuvant FOLFOX in colorectal cancer. Int. J. Cancer 132, 2209–2216 (2013).

173. Ju, H. X. et al. Distinct profiles of epigenetic evolution between colorectal cancers with and without metastasis. Am. J. Pathol. 178, 1835–1846 (2011).

174. Kim, J. C. et al. Promoter methylation of specific genes is associated with the phenotype and progression of colorectal adenocarcinomas. Ann. Surg. Oncol. 17, 1767–76 (2010).

175. Ward, R. L. et al. Adverse Prognostic Effect of Methylation in Colorectal Cancer Is Reversed by Microsatellite Instability. J. Clin. Oncol. 21 , 3729–3736 (2003).

176. Zlobec, I. et al. Stratification and Prognostic Relevance of Jass’s Molecular Classification of Colorectal Cancer. Front. Oncol. 2, 7 (2012).

177. Zanutto, S. et al. Methylation status in patients with early stage colon cancer: A new prognostic marker? Int. J. Cancer 130, 488–489 (2012).

178. Shen, L. et al. Association between DNA Methylation and Shortened Survival in Patients with Advanced Colorectal Cancer Treated with 5-Fluorouracil–Based Chemotherapy. Clin. Cancer Res. 13, 6093–6098 (2007).

179. Ogino, S. et al. CpG island methylation, response to combination chemotherapy, and patient survival in advanced microsatellite stable colorectal carcinoma. Virchows Arch. 450, 529–537 (2007).

180. Jia, M., Gao, X., Zhang, Y., Hoffmeister, M. & Brenner, H. Different definitions of CpG island methylator phenotype and outcomes of colorectal cancer: a systematic review. Clin. Epigenetics 8, 25 (2016).

181. Hughes, L. a E. et al. The CpG island methylator phenotype in colorectal cancer: progress and problems. Biochim. Biophys. Acta 1825, 77–85 (2012).

182. English, D. R. et al. Ethnicity and risk for colorectal cancers showing somatic BRAF V600E mutation or CpG island methylator phenotype. Cancer Epidemiol. Biomarkers Prev. 17, 1774–80 (2008).

183. Sjo, O. H. Prognostic factors in colon cancer. (University of Oslo, 2012). 184. Adam, R. et al. Patients with initially unresectable colorectal liver metastases: Is there a possibility

of cure? J. Clin. Oncol. 27, 1829–1835 (2009). 185. van Dijk, T. H. et al. Evaluation of short-course radiotherapy followed by neoadjuvant bevacizumab,

capecitabine, and oxaliplatin and subsequent radical surgical treatment in primary stage IV rectal cancer. Ann. Oncol. 24 , 1762–1769 (2013).

186. Kanwar, S. S., Poolla, A. & Majumdar, A. P. Regulation of colon cancer recurrence and development of therapeutic strategies. World J. Gastrointest. Pathophysiol. 3, 1–9 (2012).

187. Shima, K. et al. Prognostic Significance of CDKN2A (p16) Promoter Methylation and Loss of Expression in 902 Colorectal Cancers: Cohort Study and Literature Review. Int. J. Cancer 128, 1080–1094 (2011).

188. Xing, X. et al. The prognostic value of CDKN2A hypermethylation in colorectal cancer: a meta-analysis. Br. J. Cancer 108, 2542–8 (2013).

189. Deschoolmeester, V., Baay, M., Specenier, P., Lardon, F. & Vermorken, J. B. A review of the most promising biomarkers in colorectal cancer: one step closer to targeted therapy. Oncologist 15, 699–731 (2010).


81

APPENDIX I Association between represented variables and each of the five genes/loci constituting the classic CIMP panel. Classification of tumours according to the methylation status of each gene or locus (MINT1, MINT2, MINT31 and

CDKN2A(p16)), and its distribution and association with all represented variables. P-values were calculated using the

Chi-squared test or the Fisher’s exact test. Significant P-values (P<0.05) are represented in bold. Non-significant results

for tumour invasion depth (T), distant metastasis (M), neoadjuvant and adjuvant therapies variables excluded from the

represented table in the section “Results” are coloured in red.


M UM M UM M UM M UM


14 6.6 197 93.4 31 14.7 180 89.6 32 15.2 179 84.8 24 11.4 187 88.6

Gender Female Male

8 6

57.1 42.9

64

133

32.5 67.5

11 20

35.5 64.5

61

119

33.9 66.1

14 18

43.8 56.3

58

121

32.4 67.6

7

17

29.2 70.8

65

122

34.8 65.2

P=0.080 P=0.841 P=0.229 P=0.654

Age ≤61 >61

6 8

57.1 42.9

108 89

54.8 45.2

12 19

38.7 61.3

102 78

56.7 43.3

16 16

50.0 50.0

98 81

54.7 45.3

11 13

45.8 54.2

103 84

55.1 44.9

P=0.418 P=0.079 P=0.701 P=0.515

Location Rectum Distal Proximal

6 5 3

42.9 35.7 21.4

98 67 32

49.7 34.0 16.3

18 8 5

58.1 25.8 16.1

86 64 30

47.8 35.6 16.7

14 9 9

43.8 28.1 28.1

90 63 26

50.3 35.2 14.5

14 6 4

58.3 25.0 16.7

90 66 31

48.1 35.3 16.6

P=0.840* P=0.520 P=0.160 P=0.574

KRAS WT Mutated ND

5 8 1

38.5 61.5

-

109 75 13

59.2 40.8

-

12 17 2

41.4 58.6

-

102 66 12

60.7 39.3

-

10 20 2

33.3 66.7

-

104 63 12

62.3 37.7

-

9

12 3

42.9 57.1

-

105 71 11

59.7 40.3

-

P=0.158 P=0.067 P=0.004 P=0.164


5 8 1

38.5 61.5

-

45

151 1

23.0 77.0

-

9

22 -

29.0 71.0

-

41

137 2

22.6 77.4

-

10 22 -

31.3 68.7

-

40

137 2

22.6 77.4

-

5

19 -

24.9 75.1

-

45

140 2

24.3 75.7

-

P=0.310 P=0.496 P=0.367 P=0.804

T T1&T2 T3&T4 ND

2

11 1

15.4 84.6

-

20

176 1

10.2 89.8

-

2

29 -

6.50 93.5

-

20

158 2

11.2 88.8

-

4

28 -

12.5 87.5

-

18

159 2

10.2 89.8

-

2

22 -

8.30 91.7

-

20

165 2

10.8 89.2

-

P=0.632 P=0.542 P=0.754 P=1.000

N N0 N1/(N1&2) N2 ND

9 4 - 1

69.2 30.8

- -

66

126 - 5

34.4 65.6

- -

11 5

14 1

36.6 16.7 46.7

-

64 59 52 5

36.6 33.7 29.7

-

14 6

12 -

43.3 18.8 37.5

-

61 58 54 6

35.3 33.5 31.2

-

8 5

11 -

25.0 29.2 45.8

-

67 59 55 6

37.0 32.6 30.4

-

P=0.017 P=0.097 P=0.253 P=0.275


M

M0 M1 ND

5 8 1

38.5 61.5

-

97 99 1

49.5 50.5

-

17 14 -

54.8 45.2

-

85 93 2

47.7 52.2

-

15 17 -

46.9 53.1

-

87 90 2

49.2 50.8

-

12 12 -

50.0 50.0

-

90 95 2

48.6 51.4

-

P=0.570 P=0.560 P=0.850 P=1.000

Neoadjuvant Yes No ND

5 9 0

35.7 64.3

-

64

132 1

67.3 32.7

-

13 18 -

41.9 58.1

-

56

123 1

31.3 68.7

-

11 21 -

34.4 65.6

-

58

120 1

32.6 67.4

-

6

18 -

25.0 75.0

-

63

123 1

33.9 66.1

-

P=0.777 P=0.300 P=0.840 P=0.491

Adjuvant No Yes

2

12

14.3 85.7

40

157

20.3 79.7

7

25

26.6 77.4

35

145

19.4 80.6

7

25

21.9 78.1

35

144

19.6 80.4

3

21

12.5 87.5

39

148

20.9 70.1

P=0.741 P=0.635 P=0.811 P=0.425

APPENDIX II Distribution of represented variables for CRC patients not submitted to neoadjuvant therapy and association with CIMP status. Number of cases (and respective percentage) distributed per each category from CRC cases (N=141),

excluding tumour grade. Dichotomous (positive/negative) and trichotomous (CIMP-0, CIMP-Low and excluding CIMP-

High) CIMP categorization, and its distribution and association with all represented variables. P-values were calculated

using the Chi-squared test or the Fisher’s exact test. Significant P-values (P<0.05) are represented in bold.

Variables Cases (%) CIMP- CIMP+ CIMP-0 CIMP-L No. % No. % No. % No. % 130 92.3 11 7.7 92 64.5 46 34.1 Gender

Female Male

49 (34.8) 92 (65.2)

44 86

33.8 66.2

5 6

54.4 45.5

30 62

32.6 67.4

17 29

37.0 63.0

P=0.514 P=0.704 Age at diagnosis

≤61 >61

76 (53.9) 65 (46.1)

73 57

56.2 43.8

3 8

27.3 72.7

49 43

53.3 46.7

27 19

58.7 41.3

P=0.118 P=0.589 Tumour location

Rectum Distal Colon (Colon) Proximal Colon

50 (35.5) 63 (44.7) [91 (64.5)] 28 (19.8)

44 86 -

33.8 66.2

-

6 5 -

54.5 45.5

-

32 45 15

34.8 48.9 16.3

18 16 12

39.1 34.8 26.1

P=0.197 P=0.218 KRAS status

Wild-type Mutated ND

73 (51.8) 59 (41.8) 9 (6.4)

67 54 9

55.4 44.6

-

6 5 -

54.5 45.5

-

55 32 5

63.2 36.8

-

16 26 4

38.1 61.9

- P=1.000 P=0.009 AJCC stage

I&II III&IV ND

33 (23.4) 106 (75.2) 2 (1.4)

31 97 2

24.2 75.8

-

2 9 -

18.2 81.8

-

22 69 1

24.2 75.8

-

9

36 1

20.0 80.0

- P=1.000 P=0.668 Tumour invasion depth (T)

T1&T2 T3&T4 ND

13 (9.2) 126 (89.4) 2 (1.4)

12

116 2

9.40 90.6

-

1

10 -

9.10 90.9

-

9

82 1

9.90 90.1

-

4

41 1

8.90 91.1

-


83

P=1.000 P=1.000 Lymph node metastasis (N)

N0 N1 (N1&N2) N2 ND

51 (36.2) 39 (27.7) [87 (61.7)] 48 (34.0) 3 (2.1)

47 80 - 3

37.0 63.0

- -

4 7 - -

36.4 63.6

- -

35 32 23 2

38.9 35.6 25.5

-

13 7

25 1

28.9 15.6 55.5

- P=1.000 P=0.002 Distant Metastasis (M)

M0 M1 ND

68 (48.2) 71 (50.4) 2 (1.4)

63 65 2

49.2 50.8

-

5 6 -

45.5 54.5

-

42 49 1

46.2 53.8

-

24 21 1

53.3 46.7

- P=1.000 P=0.469 Adjuvant therapy

Yes No

113 (80.1) 28 (19.9)

103 27

79.2 20.8

10 1

90.9 9.10

72 20

78.3 21.7

39 7

84.8 15.2

P=0.693 P=0.495

APPENDIX III Association between represented variables and each of the five genes/loci constituting the classic CIMP panel for CRC patients not submitted to neoadjuvant therapy. Classification of tumours (N=141) according to the

methylation status of each gene or locus (MINT1, MINT2, MINT31 and CDKN2A(p16)), and its distribution and association

with all represented variables. P-values were calculated using the Chi-squared test or the Fisher’s exact test.


M UM M UM M UM M UM


9 6.3 132 93.7 18 12.7 123 87.3 21 14.8 120 92.2 18 12.7 123 87.3

Gender Female Male

5 4

55.6 44.4

44 88

33.3 66.7

6

12

33.3 66.7

43 80

35.0 65.0

10 11

47.6 52.4

39 81

32.5 67.5

7

11

38.9 66.1

42 81

34.1 65.9

P=0.276 P=1.000 P=0.216 P=0.795

Age ≤61 >61

3 6

33.3 66.7

73 59

55.3 44.7

7

11

38.9 61.1

69 54

56.1 43.9

12 9

57.1 42.9

64 56

53.3 46.7

8

10

44.4 55.6

68 55

55.3 44.7

P=0.302 P=0.209 P=0.815 P=0.453

Location Rectum Colon

2 7

22.2 77.8

48 84

36.4 63.6

8

10

44.4 55.6

42 81

34.1 65.9

6

15

28.6 71.4

44 76

36.7 63.3

9 9

50.0 50.0

41 82

33.3 66.7

P=0.492 P=0.435 P=0.623 P=0.192

KRAS WT Mutated ND

5 4 -

55.6 44.4

-

69 55 9

55.3 44.7

-

9 8 1

52.9 47.1

-

64 51 8

55.7 44.3

-

7

13 1

35.0 65.0

-

66 46 8

58.9 41.1

-

6

10 2

37.5 62.5

-

67 49 7

57.8 42.2

-

P=1.000 P=1.000 P=0.054 P=0.180


2 6 1

25.0 75.0

-

31

100 1

23.7 76.3

-

4

14 -

22.2 77.8

-

29 92 2

24.0 76.0

-

7

14 -

33.3 66.7

-

26 92 2

22.0 78.0

-

4

14 -

22.2 77.8

-

29 92 2

24.0 76.0

-


P=1.000 P=1.000 P=0.274 P=1.000

T T1&T2 T3&T4 ND

1 7 1

12.5 87.5

-

12

119 1

9.20 90.8

-

1

17 -

5.60 94.4

-

12

109 2

9.90 89.3

-

2

19 -

9.50 90.5

-

11

107 2

9.3

90.7 -

1

17 -

5.60 94.4

-

12

109 2

9.90 90.1

-

P=0.554 P=1.000 P=1.000 P=1.000

N N0 N1&2 ND

5 3 1

62.5 37.5

-

46 84 2

35.4 64.6

-

6

12 -

33.3 66.7

-

45 75 3

37.5 62.5

-

9

12 -

42.9 57.1

-

42 75 3

35.9 64.1

-

6

12 -

33.3 66.7

-

45 75 3

37.5 32.5

-

P=0.145 P=0.799 P=0.625 P=0.799

M M0 M1 ND

2 6 1

25.0 75.0

-

66 65 1

50.4 49.6

-

11 7 -

61.1 38.9

-

57 64 2

47.1 52.9

-

12 9 -

57.1 42.9

-

56 63 2

47.5 52.5

-

10 8 -

55.6 44.4

-

58 63 2

47.9 52.1

-

P=0.275 P=0.318 P=0.481 P=0.618

Adjuvant No Yes

1 8

11.1 88.9

27

105

20.5 79.5

3

15

16.7 83.3

25 98

20.3 79.7

5

16

23.8 76.2

23 97

19.2 81.8

2

16

11.1 88.9

26 97

21.1 78.9

P=0.688 P=1.000 P=0.568 P=0.527

APPENDIX IV Univariable and multivariable prognostic analyses: disease-specific survival analysis of CRC patients not submitted to neoadjuvant therapy according to represented variables and CIMP panel/markers methylation.

Multivariable analysis was performed considering only those variables presenting a P-value<0.05 in the univariable

analysis (excluding T, N and M stages). Significant P-values (P<0.05) are represented in bold.

Variables Univariable analysis Multivariable analysis HR (95 % CI) P HR (95 % CI) P Gender

Female (49) Male (92)

1 (referent)

0.877 (0.598-1.286)

-

0.502

Age ≤61 (76) >61 (65)

1 (referent)

1.544 (1.072-2.222)

-

0.020

1 (referent)

1.523 (1.046-2.217)

-

0.028 Location


0.789 (0.455-1.199) 0.739 (0.419-1.131)

1 (referent)

0.318 0.141 0.221

-

KRAS WT (73) Mutated (59)

1 (referent)

0.872 (0.596-1.274)

-

0.478

AJCC stage I & II (33) III (35) IV (71)

0.525 (0.332-0.830) 0.571 (0.365-0.891)

1 (referent)

0.006 0.006 0.014

-

0.481 (0.303-0.763) 0.513 (0.326-0.807)

1 (referent)

0.001 0.002 0.004

-


85

T

T1&T2 (13) T3 (115) T4 (11)

0.176 (0.074-0.416) 0.185 (0.094-0.360)

1 (referent)

<0.001 <0.001 <0.001

-

N N0 (51) N1 (39) N2 (48)

0.583 (0.382-0.892) 0.580 (0.362-0.929)

1 (referent)

0.020 0.013 0.023

-

M M0 (68) M1 (71)

1 (referent)

1.826 (1.264-2.637)

-

0.001

Grade G1&2 (119) G3 (5)

1 (referent)

2.424 (0.975-6.026)

-

0.057

Adjuvant Yes (113) No (28)

1.510 (0.948-2.406)

1 (referent)

0.083

-

CIMP Positive (11) Negative(130)

1.471 (0.790-2.741)

1 (referent)

0.224

-

CIMP CIMP-0 (136) CIMP-L (71)

1 (referent)

0.826 (0.559-1.222)

-

0.338

MINT1 M (9) UM (132)

1.141 (0.572-2.275)

1 (referent)

0.707

-

MINT2 M (18) UM (123)

0.851 (0.493-1.468)

1 (referent)

0.561

-

MINT31 M (21) UM (120)

0.674 (0.402-1.130)

1 (referent)

0.134

-

P16 M (23) UM (187)

1.791 (1.080-2.971)

1 (referent)

0.024

-

1.838 (1.090-3.097)

1 (referent)

0.022

-

APPENDIX V Univariable and multivariable prognostic analyses. Disease-specific survival (DSS) analysis of CIMP+, CIMP-0 or CIMP-L CRC patients according to other clinicopathological and molecular variables. Multivariable analysis was

performed considering only those variables presenting a P-value<0.05 in the univariable analysis (excluding T, N and M

stages). Significant P-values (P<0.05) are represented in bold.

Variables Univariable analysis Multivariable analysis

CIMP+ CIMP-0 CIMP-L

HR (95 % CI) P HR (95 % CI) P HR (95 % CI) P Gender

Female Male

1 (referent)

2.189 (0.803-5.967)

-

0.126

Age ≤61 >61

1 (referent)

1.262 (0.440-3.623)

-

0.665


Location

Rectum Distal (Colon) Proximal

0.630 (0.226-1.755)

[1 (referent)] -

0.376

- -

0.336 (0.186-0.608) 0.364 (0.203-0.653)

1 (referent)

0.001 <0.001 0.001

-

KRAS WT Mutated

1 (referent)

0.525 (0.184-1.499)

-

0.229

1 (referent)

0.414 (0.234-0.733)

-

0.002 AJCC stage

I & II III (III&IV) IV

0.357 (0.112-1.132)

[1 (referent)]

0.080

-

0.485 (0.291-0.807) 0.572 (0.359-0.910)

1 (referent)

0.008 0.005 0.018

-

T T1&T2 T3&T4

- -

- -

N N0 N1&N2

0.584 (0.458-0.922)

1 (referent)

0.282

-

M M0 M1

1 (referent)

10.86 (2.279-51.72)

-

0.003

Neoadjuvant Yes No

0.701 (0.257-1.912)

1 (referent)

0.488

-

2.068 (1.134-3.772)

1 (referent)

0.018

- Adjuvant

Yes No

- -

- -

APPENDIX VI Univariable prognostic analyses. Disease-specific survival (DSS) analysis for CRC patients with methylation of MINT2, MINT31 or CDKN2A(p16) promoters according to other clinicopathological and molecular variables. Only

those variables presenting a P-value<0.05 in the univariable analysis are represented. Significant P-values (P<0.05) are

represented in bold.

Variables MINT2 MINT31 CDKN2A(p16) HR (95 % CI) P HR (95 % CI) P HR (95 % CI) P KRAS

WT Mutated

0.358 (0.156-0.822) 1 (referent)

0.015

-

AJCC stage I & II III IV

0.896 (0.378-2.122) 0.300 (0.106-0.853)

1 (referent)

0.068 0.802 0.024

-

0.309 (0.100-0.952) 0.326 (0.102-1.037)

1 (referent)

0.061 0.041 0.058

-


87

APPENDIX VII Univariable prognostic analyses. Disease-specific survival (DSS) analysis for male patients, patients that were 61 or younger or patients with tumour located in the proximal colon, according to CIMP panel/markers methylation. Only those variables presenting a P-value<0.05 in the univariable analysis are represented.

Variables Male ≤61 Proximal colon HR (95 % CI) P HR (95 % CI) P HR (95 % CI) P CIMP

CIMP-0 CIMP-L

1 (referent) 0.369 (0.168-0.810)

-

0.013 P16

M UM

1.734 (1.020-2.946)

1 (referent)

0.042

-

2.168 (1.105-4.253)

1 (referent)

0.024

-

APPENDIX VIII Univariable prognostic analyses. Disease-specific survival (DSS) analysis for patients with KRAS WT patients, patients with mutated KRAS or patients not submitted to adjuvant therapy, according to CIMP panel/markers methylation. Only those variables presenting a P-value<0.05 in the univariable analysis are represented.

Variables KRAS WT KRAS Mutation No Adjuvant HR (95 % CI) P HR (95 % CI) P HR (95 % CI) P CIMP

CIMP-0 CIMP-L

1 (referent)

1.871 (1.159-3.021)

-

0.010

1 (referent)

0.554 (0.343-0.893)

-

0.015

MINT31 M UM

0.535 (0.305-0.938) 1 (referent)

0.029

-

0.377 (0.144-0.984)

1 (referent)

0.046

-

epigenetic study of colorectal cancer: lncrnas and cimp ... · biologia do cancro, do ipo-porto. a...

Documents